Georgetown University Law Center
Scholarship @ GEORGETOWN LAW
2009
Empirical Health Law Scholarship: The State of the
Field
Michelle M. Mello
Harvard School of Public Health
Kathryn Zeiler
Georgetown University Law Center, zeiler@law.georgetown.edu
This paper can be downloaded free of charge from:
http://scholarship.law.georgetown.edu/facpub/388
96 Geo. L.J. 649-702 (2008)
This open-access article is brought to you by the Georgetown Law Library. Posted with permission of the author.
Follow this and additional works at: http://scholarship.law.georgetown.edu/facpub
Part of the Health Law Commons
GEORGETOWN LAW
Faculty Publications
July 2009
Empirical Health Law Scholarship:
The State of the Field
96 Geo. L.J. 649-702 (2008)
Michelle M. Mello
C. Boyden Gray Associate Professor of Health
Policy and Law
Department of Health Policy and Management
Harvard School of Public Health
mmello@hsph.harvard.edu
Kathryn Zeiler
Professor of Law
Georgetown University Law Center
kmz3@law.georgetown.edu
This paper can be downloaded without charge from:
Scholarly Commons: http://scholarship.law.georgetown.edu/facpub/388/
SSRN: http://ssrn.com/abstract=1430932
Posted with permission of the author
DISEASE PREVENTION AND HEALTH
OUTCOMES
Empirical Health Law Scholarship: The State of
the Field
MICHELLE M. MELLO* & KATHRYN ZEILER**
The last three decades have seen the blossoming of the fields of health law
and empirical legal studies and their intersection—empirical scholarship in
health law and policy. Researchers in legal academia and other settings have
conducted hundreds of studies using data to estimate the effects of health law on
accident rates, health outcomes, health care utilization, and costs, as well as
other outcome variables. Yet the emerging field of empirical health law faces
significant challenges—practical, methodological, and political. The purpose of
this Article is to survey the current state of the field by describing commonly
used methods, analyzing enabling and inhibiting factors in the production and
uptake of this type of research by policymakers, and suggesting ways to increase
the production and impact of empirical health law studies. In some areas of
inquiry, high-quality research has been conducted, and the findings have been
successfully imported into policy debates and used to inform evidence-based
lawmaking. In other areas, the level of rigor has been uneven, and the best
evidence has not translated effectively into sound policy. Despite challenges and
historical shortcomings, empirical health law studies can and should have a
substantial impact on regulations designed to improve public safety, increase
both access to and quality of health care, and foster technological innovation.
TABLE OF CONTENTS
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
651
I. WHY EMPIRICISM HAS COME TO HEALTH LAW . . . . . . . . . . . . . . . .
653
A.
DATA ON OUTCOME VARIABLES
........................
653
B.
STATE-BASED LAW
.................................
654
C.
HUMAN RESOURCES
................................
655
D.
EXPANDING AUDIENCE
..............................
655
* C. Boyden Gray Associate Professor of Health Policy and Law in the Department of Health Policy
and Management of the Harvard School of Public Health. © 2008, Michelle M. Mello.
** Professor of Law at the Georgetown University Law Center. The authors thank Kevin Davis,
Daniel Fox, I. Glenn Cohen, Lawrence Gostin, and Charles Silver for helpful comments, and Isaac J.
Wheeler and Alexander Parachini for excellent research assistance. © 2008, Kathryn Zeiler.
649
650
THE GEORGETOWN LAW JOURNAL
E.
[Vol. 96:649
........................
656
II. METHODS OF EMPIRICAL HEALTH LAW RESEARCH . . . . . . . . . . . . . .
657
GROWING POLICY SIGNIFICANCE
A.
THE SPECTRUM OF ANALYTIC METHODS
...................
657
B.
MAJOR METHODOLOGICAL CHALLENGES
...................
662
III. THE IMPACT OF EMPIRICAL HEALTH LAW SCHOLARSHIP ON POLICY:
TWO CASE STUDIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
666
A.
.........................
666
1.
Graduated Driver Licensing . . . . . . . . . . . . . . . . . . . . . .
667
2.
Minimum Legal Drinking Age Laws . . . . . . . . . . . . . . .
671
3.
Enabling and Inhibiting Factors . . . . . . . . . . . . . . . . . . .
676
.......................
679
1.
The Production of Scholarship and Policy . . . . . . . . . . .
679
2.
Enabling and Inhibiting Factors . . . . . . . . . . . . . . . . . . .
687
......................................
691
IV. FUTURE DIRECTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
692
B.
C.
A.
B.
MOTOR VEHICLE SAFETY LAWS
MEDICAL MALPRACTICE REFORM
SUMMARY
WAYS TO INCREASE PRODUCTION OF HIGH-QUALITY EHL STUDIES
..
692
1.
Increase Access to and Collection of Reliable Data . . . . .
692
2.
Break Down Institutional Barriers . . . . . . . . . . . . . . . . .
693
WAYS TO INCREASE THE IMPACT OF EHL STUDIES ON HEALTH LAW
.....................................
696
1.
Strengthen Rigor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
696
2.
Increasing Widespread and Effective Dissemination of
New Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
698
Educating Consumers of EHL Research . . . . . . . . . . . . .
700
......................................
701
CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
702
AND POLICY
3.
C.
SUMMARY
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
651
INTRODUCTION
In recent years, the field of empirical legal studies has grown exponentially.1
Although empirical methods have been used to study some areas of law—
corporate law, securities regulation, antitrust, family law, and criminal law, for
example—more than others, nearly every area of law has been explored using
empirical methods. Undoubtedly, health law and policy have been among the
most widely studied areas. Examples include, but certainly are not limited to,
studies estimating the influence of statutory damages caps on medicalmalpractice insurance premiums2 and physician location decisions,3 the influence on job mobility of tax laws that encourage employers to provide health
insurance as an employee benefit,4 the impact of regulation of prices and
hospital investment on hospital investments and quality of care,5 the impact of
government regulation on tobacco consumption,6 and the effects of basing
Medicare physician payment on performance or on quality of care.7
Although the field of empirical health law (EHL) is still in its adolescence,
the increased visibility of this research and its ever-increasing policy salience
make it timely to survey the field. In this Article, we take up this task—
describing commonly used methods, analyzing enabling and inhibiting factors
in the production and uptake of EHL research by policy makers, and suggesting
ways to increase the volume, rigor, and impact of EHL studies.
We define the field of EHL to include quantitative and qualitative investigations of the effects of enacted and proposed health laws, regulations, and
policies on economic, social, and health outcomes. Two key features distinguish
methods used in EHL research from other approaches to legal scholarship:
1. See Lee Epstein et al., On the Effective Communication of the Results of Empirical Studies (pt. 1),
59 VAND. L. REV. 1811, 1816 (2006) (summarizing recent developments in the field of empirical legal
studies); Tracey E. George, An Empirical Study of Empirical Legal Scholarship: The Top Law Schools,
81 IND. L.J. 141, 142 (2006) (documenting the rise in empirical legal scholarship); James Lindgren,
Predicting the Future of Empirical Legal Studies, 86 B.U. L. REV. 1447, 1452–54 (2006) (summarizing
recent developments in the field of empirical legal studies); Richard H. McAdams & Thomas S. Ulen,
Empirical and Experimental Methods in Law, 2002 U. ILL. L. REV. 791, 791 (2002). The first annual
Conference on Empirical Legal Studies was held at the University of Texas Law School in October
2006, during which scholars presented more than eighty empirical studies.
2. For a review of the literature, see Patricia M. Danzon, Liability for Medical Malpractice, in 1B
HANDBOOK OF HEALTH ECONOMICS (A.J. Culyer & J.P. Newhouse eds., 2000).
3. See, e.g., Jonathan Klick & Thomas Stratmann, Medical Malpractice Reform and Physicians in
High Risk Specialties (Apr. 10, 2007) (unpublished manuscript, on file with authors), available at
http://www.gmu.edu/jbc/stratmann/reformDDDJLS.pdf.
4. See, e.g., Brigitte C. Madrian, Employment-Based Health Insurance and Job Mobility: Is There
Evidence of Job-Lock?, 109 Q.J. ECON. 27 (1994).
5. For a review of the EHL literature on this topic, see David S. Salkever, Regulation of Prices and
Investment in Hospitals in the United States, in HANDBOOK OF HEALTH ECONOMICS, supra note 2, at 1489.
6. See, e.g., Peter Bardsley & Nilss Olekalns, Cigarette and Tobacco Consumption: Have AntiSmoking Policies Made a Difference?, 75 ECON. REC. 225 (1999); Eugene M. Lewit et al., The Effects of
Government Regulation on Teenage Smoking, 24 J.L. & ECON. 545 (1981).
7. For a review of EHL studies on this topic, see David A. Hyman & Charles Silver, You Get What
You Pay for: Result-Based Compensation for Health Care, 58 WASH. & LEE L. REV. 1427 (2001).
652
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
utilization of data of some kind and employment of the scientific method as
opposed to legal argumentation or other forms of advocacy. Research designs
used to conduct EHL research range from pure qualitative studies to controlled
observational studies to experimental designs. We assess how well these methods are used in EHL research and discuss major methodological challenges,
including the dearth of available data, difficulties in controlling for confounding
variables, and obstacles to communicating results to policymakers.
We also assess whether and how EHL scholarship has impacted policy.8 We
present two case studies—motor vehicle safety-related laws and medical malpractice reform—that demonstrate the successful and unsuccessful application,
respectively, of social-science research to public policy. Studies of motor vehicle safety regulation provided rigorous and timely evaluations of initial
experiments with new laws relating to young drivers and identified desirable
laws. Political interest groups helped to disseminate study findings to policymakers. As a result, effective laws were adopted widely. Conversely, studies of
medical malpractice have had little effect on policy. This is partly because the
research is not uniformly of high quality, partly because the studies are too
methodologically complex for easy digestion by non-experts, and partly because some interest groups have undermined rather than promoted policymakers’ uptake of rigorous, scholarly work.
Although some barriers that hinder EHL’s impact on policy are beyond
researchers’ control, EHL scholars can take steps to increase the impact of
research findings on policy. Available steps include: attracting more funding to
obtain useful data, encouraging interdisciplinary collaborations, and training
emerging legal scholars in scientific and empirical methods. Additionally, scholars can strive to increase the rigor of EHL studies, improve dissemination
efforts, and help policymakers to become educated consumers of empirical legal
research. Our aim is to propose a roadmap for those in the academy interested in
increasing the impact of EHL scholarship and to encourage legal scholars
interested in using empirical data to influence health policy to transform their
scholarly approach, thereby moving away from the role of the expert-advocate
and toward the social-science model of scholarship.
In Part I of this Article, we offer an account of why empiricism has gained
popularity in health law and policy. In Part II we describe methodological
8. Several scholars have investigated whether and how empirical legal studies more generally have
helped shape law and policy, though they do not consider issues unique to the health law context. See,
e.g., Dean M. Hashimoto, Justice Brennan’s Use of Scientific and Empirical Evidence in Constitutional
and Administrative Law, 32 B.C. L. REV. 739 (1991) (providing examples of Justice Brennan’s use and
nonuse of empirical studies in the resolution of legal claims); John O. McGinnis, Age of the Empirical,
POL’Y REV., June–July 2006, at 47 (discussing the rise of empirical research and its potential to
influence policy); Carl E. Schneider & Lee E. Teitelbaum, Life’s Golden Tree: Empirical Scholarship
and American Law, 2006 UTAH L. REV. 53 (arguing the benefits of using empirical scholarship
generally to get the law right); Elizabeth Warren, The Market for Data: The Changing Role of Social
Sciences in Shaping the Law, 2002 WIS. L. REV. 1 (discussing the role—and sometimes non-role—of
empirical research in the effort to rewrite American bankruptcy law).
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
653
approaches and challenges. In Part III we examine enabling and inhibiting
factors in EHL research through the lens of case studies in motor vehicle safety
laws and medical malpractice reform. Finally, in Part IV, drawing on lessons
learned from the case studies, we offer suggestions for increasing EHL’s rigor,
prevalence, and impact on policy.
I. WHY EMPIRICISM HAS COME TO HEALTH LAW
Empirical scholars have investigated the connections between legal rules and
outcomes over a wide range of legal topics, but health law has been a particularly active field. This development is a positive one given that empirical
investigations of health law are a necessary component of sound policymaking.
Policy proposals often emanate from theories of how changes in policy will
impact various outcome measures or claims about the status quo (for example,
the impact of large jury awards on medical malpractice insurance premiums).
Empirical verification of the theoretical predictions, assumptions, and rhetorical
claims on which policy proposals are based increases lawmakers’ confidence in
the likely efficacy of proposals. It is critical, therefore, for empiricists to provide
information about the relationship between health law and outcomes.
Several reasons account for the popularity of health law as a topic of
investigation among empiricists. First, data produced by and gathered from
health care industries have become increasingly available along with a great
deal of administrative and survey data on health status and conditions. Second,
the dominance of state regulation in health care has generated variation and
natural experiments that can be exploited to estimate the effects of various
health laws on outcomes. Third, in recent years, law schools have hired a
greater number of faculty trained in both law and empirical methods, which has
increased the capacity for and interest in empirical legal studies generally.
Fourth, the audience interested in the findings of EHL research has expanded,
increasing the demand for it. Finally, demand has also been expanded by the
positive impact the EHL field has had on policy and by growing concern among
policymakers about health care spending and perceived crises in several corners
of the health care system, including the medical malpractice liability system and
health care insurance markets. We cover each of these in more detail below.
A. DATA ON OUTCOME VARIABLES
Despite the recent explosion in empirical legal research, many corners of the
field are plagued by a paucity of accessible data that could be used to estimate
the effects of legal rules on outcomes of interest.9 Although EHL scholars also
9. See, e.g., Gillian K. Hadfield, Judging Science: An Essay on the Unscientific Basis of Beliefs
About the Impact of Legal Rules on Science and the Need for Better Data About Law, 14 J.L. & POL’Y
137 (2006) (discussing the general dearth of data necessary to study the impact of legal rules and
policy); Michael Heise, The Importance of Being Empirical, 26 PEPP. L. REV. 807, 822 (1999) (pointing
to underinvestment in data production as the reason for the dearth of data).
654
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
grapple with difficulties in obtaining data useful for advancing our knowledge
of the connections between health law and outcomes,10 there has been relatively
good access, in some cases for decades, to various datasets on health care
utilization and health outcomes.11 Regulators continue to open useful datasets to
the research community,12 and researchers have often convinced market actors
to do the same. The data usually come with strict confidentiality agreements,
limiting the potential for other research teams to test the validity and robustness
of the reported results, but the data have allowed researchers to explore interactions between markets and legal rules that formerly resided in black boxes.
B. STATE-BASED LAW
A second competitive advantage for EHL researchers is that states have been
the primary loci of lawmaking around important aspects of health care markets,
including insurance coverage, regulation of health care delivery, and tort law.13
The state-based nature of legal rules confers two specific advantages for research. First, adoption of legal reforms generally happens in waves rather than
in all states at once. This pattern provides the cross-sectional and longitudinal
variation researchers need for robust analysis of the effects of changes in legal
rules on health outcomes. Second, that health care markets are made up of a
large number of moving pieces increases the likelihood that legal rules are
adopted for reasons unrelated to the outcome variable of interest—an important
precondition for empirical models to produce unbiased estimates of the effects
of the rules on the outcome variable.14
10. See infra Part IV.
11. Examples include the Healthcare Cost and Utilization Project (available since 1988), the
Longitudinal Study of Aging (available since 1984), and the National Health Interview Survey
(available since 1957). On the other hand, developments in privacy law have made it more difficult for
researchers to access individual patients’ medical records for purposes of doing chart review studies
(the Harvard Medical Practice Study, see infra note 60, is an example of such a study).
12. Examples of datasets that have more recently become available online include the National
Practitioner Data Bank (reporting data on medical malpractice payments and various sanctions against
practicing physicians since 1990), the Dartmouth Atlas of Health Care (reporting on hospital expenditures and costs, hospital capacity, physician and specialist work forces, frequency of diagnostic and
surgical procedures, and Medicare payments for care provided by hospitals since 1996), and the
Employer Health Benefits Survey (providing data on employer-sponsored health benefits since 1998).
13. Examples of legal rules related to health care adopted at the state level include: tort reform and
legal standards of care enforced by tort law; safety regulations such as seatbelt and workplace smoking
laws; information disclosure regulations such as mandated disclosure of provisions of contracts
between managed care organizations and physicians; regulation of health insurance coverage; taxation
and other forms of regulation related to vices such as tobacco and alcohol; and regulation of access to
unhealthful foods and beverages in schools. Of course, empirical researchers in other areas of law,
including criminal law, contract law, and property law, enjoy similar benefits bestowed by gradual and
partial adoption of reforms by states.
14. When there is a two-way relationship between the legal reform adopted and the outcome it aims
to improve—for example, states with high malpractice insurance premiums are more likely to adopt
malpractice damages caps, and damages caps have the effect of reducing premiums—a problem of
endogeneity results, necessitating the use of sophisticated (and sometimes infeasible) analytical meth-
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
655
C. HUMAN RESOURCES
Historically, the bulk of EHL research has been conducted outside law
schools, in public health and public policy schools, business schools, and social
science departments—including economics, political science, and psychology
departments. This is unsurprising in light of the fact that this type of research
arises out of scientific and social-scientific traditions.
In recent years, however, we have witnessed an increase in EHL research
performed within the legal academy and published in law journals. Several
factors may account for this. Pioneering work by “bridge” scholars, such as Dan
Kessler and Troy Brennan, who are dually trained in law and empirical methods, has been important in establishing prominence and respectability for this
research within the legal academy.15 Health economists’ interest in regulatory
questions—propelled by Kenneth Arrow’s influential article,16 which helped
shape health economics as a field—spurred recognition of potential synergies of
interdisciplinary collaborations between legal scholars and economists. Interest
in these same health care regulatory questions within the legal academy has led
to the creation of interdisciplinary health law research programs at leading
universities, including Georgetown/Johns Hopkins and Harvard, which provide
dedicated institutional support for collaborative, empirical research. Finally,
carried by this momentum, a number of law schools recently have appointed
faculty members trained in both law and empirical methodology and interested
in EHL research.17
D. EXPANDING AUDIENCE
The move to increase production capacity for EHL research has been fueled
by an increasing demand for this type of scholarship in several quarters.
Physicians and institutional health care providers, who highly value quantitative
research and evidence-based decisionmaking, are regular consumers of EHL
research as a source of information about how the law potentially impacts their
practice and industry. This has recently become true of other health-care-market
actors as well, including medical malpractice insurers, health insurers, employers who provide health insurance as a benefit to their employees, and government agencies that finance health care purchases for subsets of the population.
ods. For a more technical explanation, see JEFFREY M. WOOLDRIDGE, INTRODUCTORY ECONOMETRICS: A
MODERN APPROACH 521–29 (3d ed. 2006).
15. See, e.g., Troyen A. Brennan et al., Incidence of Adverse Events and Negligence in Hospitalized
Patients: Results of the Harvard Medical Practice Study I, 324 NEW ENG. J. MED. 370 (1991); Daniel P.
Kessler & Mark B. McClellan, Do Doctors Practice Defensive Medicine?, 111 Q.J. ECON. 353 (1996).
16. Kenneth J. Arrow, Uncertainty and the Welfare Economics of Medical Care, 53 AM. ECON. REV.
941 (1963).
17. In addition to one of the authors, recent hires include Jill Horwitz (University of Michigan Law
School), Anup Malani (University of Chicago Law School), Jonathan Klick (Florida State University
Law School), Ronen Avraham (Northwestern University Law School), and Daniel Ho (Stanford Law
School).
656
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
Public health communities also are regular consumers of EHL studies. These
stakeholder groups value studies focused on potential legal solutions to, and
sometimes law-related causes of, public health problems (for example, obesity,
tobacco- and alcohol-related illnesses, and limited access to care). In addition,
EHL studies focused on the impact of various legal institutions on the health
and safety of at-risk populations are of interest to public health organizations
(for example, studies shedding light on how tort reform impacts subsets of
potential victims of medical malpractice). Empirical studies of legal rules
intended to control the spread of infectious diseases, including compulsory
vaccination and compulsory treatment, are also of special interest.
Legal system insiders, including advocates, judges, and legislators, have also
increased their demand for EHL research. More than ever before, lawyers rely
on empirical studies to support legal arguments made to courts, arbiters, and
administrative agencies in health-related matters. For example, the Supreme
Court brief filed by the respondents in Ashcroft v. Raich (later known as
Gonzales v. Raich) cited a General Accounting Office study of the experience
with medical marijuana laws in four states.18 Likewise, judges themselves refer
to empirical research to justify their holdings. For example, in another recent
Supreme Court decision, Justice Souter cited an empirical study to support
claims regarding the influence of managed-care contracts with physicians on
physician satisfaction.19 Legislators also find these studies instructive for shaping policy and useful as ammunition to support their positions.20
E. GROWING POLICY SIGNIFICANCE
In 2005, the United States spent $2 trillion on health care, an average of
$6700 per person and 16% of the country’s Gross Domestic Product (compared
to 5% in 1960, 12% in 1990, and 13% in 2000).21 Most ($1.7 trillion) of this
spending is devoted to personal health care costs. The federal government
financed 34% of personal health expenditures and state and local governments
covered 11%.22 The sheer size of the industry and the growing burden on
taxpayers have ignited public demand for policymakers to take affirmative steps
18. Brief for Respondents at 17–18, Gonzales v. Raich, 545 U.S. 1 (2005) (No. 03-1454) (citing U.S.
GEN. ACCOUNTING OFFICE, MARIJUANA: EARLY EXPERIENCES WITH FOUR STATES’ LAWS THAT ALLOW USE
FOR MEDICAL PURPOSES 21 (2002)).
19. Pegram v. Herdrich, 530 U.S. 211, 221 (2000) (citing Grumbach et al., Primary Care Physicians’ Experience of Financial Incentives in Managed-Care Systems, 339 NEW ENG. J. MED. 1516
(1998)).
20. See, for example, CONG. BUDGET OFFICE, THE PROPOSED TOBACCO SETTLEMENT: ISSUES FROM A
FEDERAL PERSPECTIVE 4 (1998), microformed on CIS No. 98-J932-24 (Cong. Info. Serv.), which was
written at the request of the then-Senate Assistant Majority Leader and reviews the empirical literature
on the connections between cigarette price increases and tobacco consumption.
21. NAT’L CTR. FOR HEALTH STATISTICS, U.S. DEP’T OF HEALTH & HUMAN SERVS., HEALTH, UNITED
STATES, 2007: WITH CHARTBOOK ON TRENDS IN THE HEALTH OF AMERICANS 26 (2007), available at
http://www.cdc.gov/nchs/data/hus/hus07.pdf.
22. Id. at 26–27.
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
657
to ease this burden and fix other perceived health care market problems. The
availability of data, advances in techniques for estimating the effects of legal
interventions, and public intolerance for policy failures in this critical area have
pushed policymaking toward a more evidence-based approach.
A number of political interest groups have responded by churning out their
own empirical research in an effort to sway the political process in their favor.
In many cases, the research is simplistic and, by academic standards, of poor
methodological quality. This has raised the stakes for EHL scholars to contribute a better alternative.
These developments illustrate the potential for growth and impact in the field
of EHL. We further explore the status of the field by considering the state of the
methods used to conduct EHL research and the uptake of research by the policy
community as illustrated by two case studies.
II. METHODS OF EMPIRICAL HEALTH LAW RESEARCH
The methods used in EHL research are as diverse as the topics investigated.
In this Part, we discuss major methodological approaches and their relative
representation in the field and then examine some of the significant challenges
that empirical legal scholars confront in researching health law issues.
A. THE SPECTRUM OF ANALYTIC METHODS
A wide spectrum of methods has been used to empirically study developments in health law. We review the most common methodologies, ordered from
least to most scientifically rigorous, below.23
First, case studies occasionally have been used to examine health law problems, particularly problems that are unique or for which examination of extreme
examples is of interest.24 Because case studies are not controlled analyses and
the example selected for study may be unrepresentative of the broader population, case studies have limited generalizability.25 They can, however, provide a
rich, vivid, and detailed lens on a particular example of a problem or policy
innovation. A well-known example is the fascinating chronicle of the federal
government’s response to a perceived threat of epidemic swine flu in 1977,
23. This is not an exhaustive list. We do not include analytical methods that are extremely rare in
health law research, for example, epidemiological research designs.
24. See, e.g., ROGAN KERSH, THE POLITICS OF MEDICAL MALPRACTICE IN PENNSYLVANIA, 1975–2005
(2005, on file with authors) (examining a state that has suffered extreme difficulties in its malpractice
insurance market); Michelle M. Mello & Troyen A. Brennan, The Controversy over High-Dose
Chemotherapy with Autologous Bone Marrow Transplant for Breast Cancer, 20 HEALTH AFF. 101
(2001) (discussing the role of mandated-benefit laws and judicial decisions in forcing insurers to cover
an experimental therapy for breast cancer that was later shown to be ineffective). By “case study,” we
do not mean anecdotes strategically selected to support a foregone conclusion. See David A. Hyman,
Lies, Damned Lies, and Narrative, 73 IND. L.J. 797 (1998) (criticizing the use of anecdotes in health
care policymaking).
25. Arend Lijphart, The Comparable-Cases Strategy in Comparative Research, 8 COMP. POL. STUD.
158, 160 (1975).
658
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
including how it addressed vaccine manufacturers’ liability concerns.26
A second, quite prevalent form of analysis in EHL research is qualitative
analysis of judicial opinions.27 This approach bears some resemblance to traditional legal scholarship, which focuses on deep doctrinal analysis of a small
number of judicial decisions. It differs in two key respects: expansion of the
sample to include a larger number of cases and use of systematic methods of
content analysis. A typical research design involves random selection of a
sample of cases meeting certain subject-matter criteria, application of a detailed
coding and content-analysis methodology, and drawing of inferences based on
both thematic analysis and frequency counts of various case characteristics.28
This design offers a broader and more systematic examination of judicial
decisionmaking than more traditional forms of doctrinal analysis, and is very
useful for getting the “lay of the land” in an area of litigation that has not yet
been broadly characterized. Like all studies that rely on published judicial
decisions, it is limited by the potential unrepresentativeness of published decisions.29 Content analysis has also been criticized for being too subjective.30
Another application of qualitative methods in health law research is the
collection of survey and interview data used to examine the impacts of health
laws or litigation. In addition to performing content analysis, such studies may
analyze survey and interview data quantitatively. Major challenges that compromise the quality of these studies are the difficulties in avoiding selection bias
and in obtaining an adequate response rate and concerns about respondents’
self-interest in the outcome of the study.
Occasionally, researchers have conducted descriptive studies in health law
involving quantitative analysis of data from outside the legal system. The aim
typically is to contribute suggestive evidence about the impact of a law or to test
assumptions or claims policymakers and other researchers rely on to argue a
position or develop theories. Simple descriptive statistics such as the mean and
standard deviation, or frequency and proportion, are presented and may be
26. See RICHARD E. NEUSTADT & HARVEY V. FINEBERG, THE SWINE FLU AFFAIR: DECISION-MAKING ON A
SLIPPERY DISEASE (1978), cited in DAVID C. HOAGLIN ET AL., DATA FOR DECISIONS: INFORMATION STRATEGIES FOR POLICYMAKERS 120–23 (1982).
27. See Heise, supra note 9, at 824 (“Much of the existing empirical legal scholarship falls loosely
into one or more of three broad categories: judicial opinion coding or case content analysis, descriptive,
and inferential.”).
28. See Kevin M. Clermont & Theodore Eisenberg, Litigation Realities, 88 CORNELL L. REV. 119,
125–26 (2002); Mark A. Hall & Ronald F. Wright, Systematic Content Analysis of Judicial Opinions 11
(Wake Forest Univ. Legal Studies, Working Paper No. 913336, 2006), available at http://ssrn.com/
abstract⫽913336.
29. Clermont & Eisenberg, supra note 28, at 125–26 (discussing several ways in which published
decisions may be unrepresentative of the broader population of judicial decisions); William M. Sage,
Judicial Opinions Involving Health Insurance Coverage: Trompe L’Oiel or Window on the World?, 31
IND. L. REV. 49, 61–68 (1998) (describing a number of limitations of content analyses of judicial
decisions).
30. See, e.g., Heise, supra note 9, at 825.
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
659
supplemented by bivariate analyses estimating associations between variables.31
Univariate analyses have no explanatory power but can provide useful information on the nature of a health law problem or be used to support or call into
question conventional wisdom—for example, the extent to which physicians are
personally exposed to malpractice settlements in excess of their policy limits.32
The major weakness of bivariate studies is that, although some analyses may be
stratified,33 they otherwise do not control for confounding variables.34 This
limits their usefulness in supporting inferences about causal connections between the variables. Therefore, although these methods can produce very useful
information for policymakers and theorists, if limitations are not fully and
clearly disclosed, readers may incorrectly attribute observed differences between comparison groups to the variables studied, when in reality other factors
drive the differences.
More sophisticated designs—controlled observational or quasi-experimental
studies using multivariate regression techniques—constitute the bulk of empirical legal studies conducted by law-and-economics scholars.35 Such studies
attempt to overcome the problems with drawing causal inferences from univariate and bivariate analyses by employing techniques to control for confounding
variables. The simplest variant is the pre/post analysis, sometimes called an
31. Bivariate analyses involve two variables and typically consist of cross-tabulations comparing the
relationship between an outcome variable and an explanatory variable—for example, rates of mammography in states that do and do not mandate that insurers cover routine mammography. Univariate
analyses provide summary statistics about a single variable—for example, mean and variance of
mammography rates across all states.
32. See Charles Silver et al., Malpractice Payouts and Malpractice Insurance: Evidence from Texas
Closed Claims, 1990–2003, GENEVA PAPERS ON RISK & INS.: ISSUES & PRAC. (forthcoming Apr. 2008),
available at http://ssrn.com/abstract⫽983199; Kathryn Zeiler et al., Physicians’ Insurance Limits and
Malpractice Payouts: Evidence from Texas Closed Claims, 1990–2003, J. LEGAL STUD. (forthcoming)
(on file with authors).
33. A stratified analysis is one that divides observations into groups, or strata, based on their value of
some potentially confounding variable and then performs bivariate analyses within each stratum. For
example, a researcher might be interested in the relationship between state mandates for physical
education in schools and the prevalence of overweight and obesity among schoolchildren, but suspect
that the effect will be different for girls and boys. A simple way to investigate that hypothesis would be
to stratify by sex and analyze the relationship within each stratum.
34. If an analysis does not control for confounding variables, then variation in an outcome variable
may be incorrectly attributed to the explanatory variable of interest. For example, tort reforms are often
implemented as packages that include several different types of reforms (for example, damages caps
plus joint-and-several liability reform). If joint-and-several liability reforms are not included in the
analysis, observed reductions in the costs of malpractice claims might appear to be due to damages caps
when, in fact, they may be caused in part by joint-and-several-liability reform.
35. Thomas E. Willging, Past and Potential Uses of Empirical Research in Civil Rulemaking, 77
NOTRE DAME L. REV. 1121, 1132 (2002). Quasi-experimental studies involve observing some variable
of interest before and after a change that affects the entire population. If the change affects only a subset
of the population, each subject in the affected group can be matched to a subject in the non-affected
group based on important characteristics to allow for comparison of differences in some variable of
interest before and after the change across groups (this is sometimes referred to as “difference-indifference”). Alternatively, multivariate regression analysis can be used to control for potential confounding variables.
660
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
event study, which employs no explicit controls but simply measures the
outcome variable before and after the intervention. This design’s major weakness is that it does not control for exogenous changes in the environment over
time that may introduce confounding variables.36
Alternatively, multivariate regression techniques may be applied in crosssectional or time-series analyses.37 Longitudinal designs may be employed to
maximize the amount of information that can be gleaned from a limited number
of cross-sectional units (such as states) and to analyze changes over time. These
studies take advantage of variation in legal rule adoption and outcomes over
units of observation and time and implement controls to rule out alternative
explanations, allowing the researchers to conclude, with some level of confidence, that the findings either support or reject claims and theories positing
causal connections. While these studies are useful as evidence for or against
particular theories of causal connections, they should not be viewed as conclusive or touted as such.
Virtually absent from the EHL literature are experimental designs. Randomized experiments are the “gold standard” for testing the effects of interventions;38 the use of random treatment assignment is ideal for reducing the
influence of confounding variables.39 Although experimental designs have occasionally been used to test the effect of changes in health care organizational
policy that could be implemented as public policy—for example, imposing
copayments on enrollees in a private or public insurance program40—for obvious reasons they are almost never an option for evaluating the effects of health
laws.41
In evaluating the rigor of EHL scholarship as a whole, it is useful to consider
evaluative tools that reviewing organizations use to judge the scientific rigor of
36. See M. Eccles et al., Research Designs for Studies Evaluating the Effectiveness of Change and
Improvement Strategies, 12 QUALITY & SAFETY HEALTH CARE 47, 50 (2003).
37. Cross-sectional designs measure many units (for example, health care consumption for all fifty
states) at one point in time, while time-series data captures just one unit (for example, health care
consumption in Texas) at different points in time. For an in-depth discussion of experimental design,
see generally THOMAS D. COOK & DONALD T. CAMPBELL, QUASI-EXPERIMENTATION: DESIGN AND ANALYSIS
ISSUES FOR FIELD SETTINGS (1979).
38. See Eccles et al., supra note 36, at 48; see also A.L. COCHRANE, EFFECTIVENESS AND EFFICIENCY:
RANDOM REFLECTIONS ON HEALTH SERVICES 2, 22 (1972); Rachel Croson, Why and How To Experiment:
Methodologies from Experimental Economics, 2002 U. ILL. L. REV. 921, 945 (“Experiments can
accomplish objectives that other forms of analyses cannot. They can offer clean tests of economic
theories by constructing experiments that meet the assumptions of the theories, and observing the
outcomes. They can investigate alternative causes of observed anomalies . . . by independently manipulating factors that in reality are confounded.”).
39. See Eccles et al., supra note 36, at 48; Willging, supra note 35, at 1127.
40. See Willard G. Manning et al., A Controlled Trial of the Effect of a Prepaid Group Practice on
Use of Services, 310 NEW ENG. J. MED. 1505 (1984).
41. Notable exceptions exist. See, e.g., Lauren A. McCormack et al., Health Plan Decision Making
with New Medicare Information Materials, 36 HEALTH SERVICES RES. 531 (2001) (reporting results from
controlled experiments to determine the impact of Medicare informational materials on new and
experienced Medicare beneficiaries).
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
661
various study designs. Among the most prominent typologies of research designs is that of the U.S. Preventive Services Task Force.42 In considering
whether an adequate evidence base exists to support recommending a particular
preventive health measure, the Task Force applies the following rating system,
which flows from strongest to least strong research design:
I Evidence obtained from at least one properly randomized controlled trial.
II-1 Evidence obtained from well-designed controlled trials without randomization.43
II-2 Evidence obtained from well-designed cohort or case-control analytic studies, preferably from more than one center or research group.44
II-3 Evidence obtained from multiple time series with or without the intervention . . . .45
III Opinions of respected authorities, based on clinical experience, descriptive
studies and case reports, or reports of expert committees.46
This typology emphasizes the importance of randomized studies, places little
weight on descriptive studies, and views quasi-experimental designs as a heterogeneous group with some designs superior to others. To our knowledge, the
overwhelming bulk of EHL research conducted to date can best be classified as
category “II-3” or “III” studies. In addition to the paucity of controlled trials,
another notable gap in the EHL literature is the scant use of techniques of
meta-analysis or systematic review. Meta-analysis uses sophisticated statistical
methods to aggregate and collectively analyze data from multiple studies on the
same topic using similar designs. Systematic reviews apply a structured process
to compare and evaluate the relative strength of findings from a number of
42. See generally Russell P. Harris et al., Current Methods of the U.S. Preventive Servs. Task Force:
A Review of the Process, 20 AM. J. PREVENTIVE MED. 21 (2001).
43. Controlled trials without randomization are experimental studies that compare a treatment group
to a control group, but the groups are formed based on subject characteristics rather than random
assignment. The experimenter, however, maintains control over the application of the treatment. The
analysis must control for confounding effects potentially arising from self-selection into groups or other
sources of difference between groups.
44. Cohort and case-control study designs are observational, not experimental, designs. Both allow
the researcher to examine the association between one or more risk factors and an outcome variable. In
a cohort study, the researcher identifies a group (cohort) of subjects based on their exposure to some
risk factor and then observes their experience over time to see which subjects develop the outcome
condition of interest (for example, a disease). RAYMOND S. GREENBERG ET AL., MEDICAL EPIDEMIOLOGY
106 (2d ed. 1996). In a case-control study, the researcher selects a group of subjects (cases) who have a
particular condition, such as a disease. The researcher then selects a control group consisting of subjects
from the same population who do not have the condition and examines the ways in which they differ on
various risk factors. Id. at 120–21.
45. “With intervention” indicates that the researcher controls the application of the treatment over
time.
46. Rating system quoted from Harris et al., supra note 42, at 26 tbl.2.
662
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
studies examining the same phenomenon. These approaches are not without
methodological controversy,47 but they have the advantage of harnessing the
power of multiple studies. They are rarely used, however, to study the effects of
legal rules on outcomes.48 As with all empirical studies, meta-analyses and
research syntheses have limitations, but they can help to provide broader views
of treatment effects identified in different studies especially when the studies
report divergent results. Overall, the field of EHL lags well behind the standard
of evidence that researchers and policymakers aim for in other areas of evaluation of health interventions.49
B. MAJOR METHODOLOGICAL CHALLENGES
Empirical health law studies confront several significant methodological
challenges that may explain why the field has not advanced farther. Three issues
that cut across many areas of inquiry within health law are particularly troubling.
First and foremost is the problem of data availability. Although there are
excellent datasets on health care utilization and many health outcomes, the
investment that government and private research sponsors have made in building health datasets for research purposes is not mirrored in the realm of health
law. The government’s most significant commitment to empirical legal research,
the Bureau of Justice Statistics, has made only occasional forays into health law,
all limited to descriptive studies of medical malpractice claims50 and relying on
data aggregated from other sources. There are few examples of governmentfunded or government-run projects aimed at building databases of health laws.51
The best known recent instance may be the National Science Foundation’s grant
to Ronen Avraham at Northwestern University Law School to create a comprehensive database of state malpractice tort reform laws.52 Some private organizations, most notably the National Conference of State Legislatures, maintain
47. See, e.g., Peter C. Gotzsche et al., Data Extraction Errors in Meta-analyses That Use Standardized Mean Differences, 298 JAMA 430 (2007).
48. Recently published studies suggest that these approaches may be gaining traction in EHL
research, though. See, e.g., Caroline M. Fichtenberg & Stanton A. Glantz, Effect of Smoke-Free
Workplaces on Smoking Behaviour: Systematic Review, 325 BRITISH MED. J. 188 (2002) (systematically
reviewing primary empirical studies to compare the efficacy of workplace smoking bans and tax
increases); Stephen Kisely et al., Randomized and Non-randomized Evidence for the Effect of Compulsory Community and Involuntary Out-Patient Treatment on Health Service Use: Systematic Review and
Meta-analysis, 37 PSYCHOL. MED. 3 (2006).
49. Of course, the explanation rests in part on the difficulties with conducting experimental studies
in health law, as we discuss further in Part II.B.
50. See generally Bureau of Justice Statistics Publications, http://www.ojp.usdoj.gov/bjs/
pubalp2.htm#M (last visited Aug. 20, 2007).
51. This is a particular challenge for EHL researchers attempting to identify the impact of tort
reforms on medical malpractice insurance premiums and other relevant outcome variables. See infra
Part III.B.
52. See Ronen Avraham, Database of State Tort Law Reforms (Nw. Sch. of Law. & Econ., Research
Paper No. 06-08, 2006), available at http://ssrn.com/abstract⫽902711.
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
663
useful lists of pending and enacted state laws on health topics,53 but these are
not organized into a database, nor do they provide detailed information about
the laws.
Without databases custom-built for research, EHL scholars must rely on
administrative data collected for other purposes or build their own datasets from
the ground up. For research topics related to health or liability insurance,
administrative data held by private insurers, state departments of insurance,
state medical boards, and state health departments are attractive options. Unfortunately, such databases can be difficult to access; states and insurers hold them
tightly54 due to a combination of legal concerns55 and anxieties about the
potential political fallout from adverse study findings. Additionally, datacollection processes and data fields often are not standardized across states.
Single-state datasets are rarely nationally generalizable, and datasets from
multiple states cannot easily be aggregated to support national studies.
In the medical liability realm, the National Practitioner Data Bank and jury
verdicts databases offer two alternatives, though each captures only a portion of
all filed claims and includes a limited number of variables. The Data Bank most
likely suffers from underreporting56 and restricts the data fields the public may
access.57 Jury verdict reports have also been criticized for their unrepresentativeness and incompleteness.58
Administrative datasets also have the obvious shortcoming of not having
been constructed with researchers’ needs and objectives in mind. They may
omit data fields that researchers view as crucial to a rigorous analysis, and the
data may not have been collected with the degree of completeness and accuracy
that researchers desire.59 Finally, researchers external to the organization that
53. These lists are collected at National Conference of State Legislatures, NCSL 50-State Legislative
Tracking Resources, http://www.ncsl.org/programs/lis/lrl/50statetracking.htm (last visited Oct. 22, 2007).
54. See NEIL VIDMAR, MEDICAL MALPRACTICE AND THE TORT SYSTEM IN ILLINOIS: A REPORT TO THE
ILLINOIS STATE BAR ASSOCIATION 5–6 (2005), available at, http://eprints.law.duke.edu/archive/00001125
(click on PDF) (reporting that the Illinois Department of Insurance refused to grant permission to
analyze its claims data); Sage, supra note 29, at 59 (“[H]ealth insurers often assert proprietary interests
in their coverage standards.”).
55. See Michelle M. Mello et al., Market Watch: Changes in Physician Supply and Scope of Practice
During a Malpractice Crisis: Evidence from Pennsylvania, 26 HEALTH AFF. w425, w433 (2007)
(describing an instance where an insurance department expressed concerns about violating its confidentiality obligations under state law).
56. See U.S. GEN. ACCOUNTING OFFICE, NATIONAL PRACTITIONER DATA BANK: MAJOR IMPROVEMENTS
ARE NEEDED TO ENHANCE DATA BANK’S RELIABILITY 10–13 (2000), available at http://www.gao.gov/
new.items/d01130.pdf.
57. U.S. Department of Health and Human Services, Practitioner Data Banks Branch, Public Data
Files, http://www.npdb-hipdb.com/publicdata.html (last visited Oct. 22, 2007) (noting that the Data
Bank omits certain information).
58. See, e.g., Neil Vidmar, Pap and Circumstance: What Jury Verdict Statistics Can Tell Us About
Jury Behavior and the Tort System, 28 SUFFOLK U. L. REV. 1205, 1210–11, 1225–28 (1994).
59. For a discussion of the ways in which databases maintained by health insurers fall short of
providing sufficient information to evaluate medical appropriateness decisions, see, for example,
Maxwell J. Mehlman, Getting a Handle on Coverage Decisions: If Not Case Law, Then What?
Comments on a Paper by Professor William Sage, 31 IND. L. REV. 75, 75–78 (1998). On the other hand,
664
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
collected the data may find it challenging to become familiar with the dataset
and its quirks and quality problems. Unless an organizational insider is willing
to spend the time to brief the researchers on the dataset and respond to their
queries, important data-quality issues and definitional issues may go unnoticed.
Engaging in primary data collection to build new datasets is a possibility in
many areas of empirical health law research, but it can be prohibitively expensive and few extramural research sponsors are interested in funding such
enterprises. Additionally, projects that require on-the-ground data collection (for
example, reviewing medical charts to identify instances of malpractice or
reviewing insurance claims to identify wrongful denials) require substantial
infrastructure to train, house, and support a large data-collection staff. Few law
schools can afford to meet these demands, so scholars tend to rely on readily
available but very limited administrative datasets and what they can cull from
judicial opinions and electronic databases of statutes and regulations. Datacollection efforts are more common in schools of public health and medicine,60
but their heavy reliance on extramural funding can make it difficult for faculty
to pursue such projects if grants are not available to support them.
The second major methodological challenge is the difficulty of controlling for
confounding variables—a problem common to all observational studies.61 As
previously discussed, the ideal method to implement such controls is through
the use of experimental designs including randomized studies. Although there
has been some use of experimental designs in other areas of empirical legal
research focusing on judicial processes,62 there are few opportunities to conduct
pure experimental studies in health law. Judicial processes are of less interest,
and the agendas and actions of state legislatures cannot be dictated by a
randomization process. Thus, as in other areas of social science,63 researchers
must rely on carefully controlled quasi-experimental designs.
administrative data may be more accurate than data collected for research purposes because the
information providers may have stronger incentives to treat the task seriously and provide truthful
responses. HOAGLIN ET AL., supra note 26, at 159.
60. The most famous example may be the Harvard Medical Practice Study, a review of approximately 30,000 medical records that measured the incidence of medical negligence and malpractice
claiming. The main study findings are collected at PAUL C. WEILER ET AL., A MEASURE OF MALPRACTICE:
MEDICAL INJURY, MALPRACTICE LITIGATION, AND PATIENT COMPENSATION 42–59, 68–76, 92–109 (1993).
61. See generally U.S. DEP’T OF HEALTH & HUMAN SERVS., FACT SHEET: RATING THE STRENGTH OF
SCIENTIFIC RESEARCH FINDINGS (2002), http://www.ahrq.gov/clinic/epcsums/strenfact.pdf. For a general
discussion of the threats to internal validity in quasi-experimental studies, see COOK & CAMPBELL, supra
note 37, at 50–59.
62. See, e.g., Croson, supra note 38, at 935 (describing experiments that tested the effect of a
damages cap law on pretrial bargaining among parties in a hypothetical personal injury case); Willging,
supra note 35, at 1128–29 (describing an experiment with mandatory attorney conference prior to
briefing and arguing appeals in the U.S. Court of Appeals for the Second Circuit).
63. Among the barriers to randomized experiments in social-science research generally are ethical
concerns about randomization in some circumstances, differences between experimental and actual
delivery of the intervention, time and money costs, and controlling the integrity of the experiment in the
field. See PETER H. ROSSI & HOWARD E. FREEMAN, EVALUATION: A SYSTEMATIC APPROACH 290–93 (5th ed.
1993) (describing limitations on the use of randomized experiments in evaluating “social programs”).
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
665
Controlling for confounding variables is particularly hard in studies of innovations in state law, due to the profound and multifaceted cross-sectional variation. Finding appropriate data to control for variables that might be correlated
with state enactment of health laws can be a difficult, and sometimes impossible, task. Fixed-effects models offer a possible solution,64 but only where the
laws in question exhibit a reasonable amount of variation over the study period,
the number of cross-sections (for example, states) is small relative to the
number of time periods covered by the data, and the unobservable heterogeneity
is constant over time.65 Difference-in-difference models are more attractive in
some cases,66 but have rarely been used in health law studies,67 and some
researchers who have used them do not appear to have incorporated appropriate
adjustments for the inconsistency of standard errors that arises from serially
correlated data.68 Due to these complexities, there is a substantial risk of
omitted variable bias in studies of many kinds of health laws. By properly
controlling for this bias, however, parameter estimates can be used to draw
causal inferences.
The third major challenge is how to model state laws that may have similar
aims and general approaches but exhibit considerable heterogeneity in their
specific provisions. For example, some laws that aim to ensure that women will
be protected from “drive-through” breast cancer surgeries set a specific minimum hospital stay, while others provide that the length of stay shall be
determined by the woman and her physician rather than her insurance company.69 There is no consensus within the field about whether such diverse laws
should be modeled using a single dummy variable (indicating whether a law of
64. Fixed-effects models are used to control for unobservable heterogeneity that is constant over
time. See CHENG HSIAO, ANALYSIS OF PANEL DATA 30–33 (2d ed. 2003).
65. Given the limited time horizon of many studies, this often will not be the case. For example,
medical malpractice tort reforms in the states were largely time-invariant between the late 1980s and
the early 2000s. Y. Tony Yang et al., A Longitudinal Analysis of the Impact of Liability Pressure on the
Supply of Obstetrician-Gynecologists, J. EMPIRICAL LEGAL STUD. (forthcoming 2008).
66. Difference-in-difference models are used to distinguish effects caused by an “exogenous shock”
such as a legal reform from differences attributable to time trends driven by other causes.
67. Among the rare instances in which such models have been used are William E. Encinosa & Fred
J. Hellinger, Have State Caps on Malpractice Awards Increased the Supply of Physicians?, 24 HEALTH
AFF. w5-250 (2005); Kessler & McClellan, supra note 15; Daniel P. Kessler et al., Impact of
Malpractice Reforms on the Supply of Physician Services, 293 JAMA 2618 (2005); Klick & Stratmann,
supra note 3.
68. Year-to-year correlations in the dependent variable may bias standard errors downward, resulting
in overestimation of significance levels. Marianne Bertrand et al., How Much Should We Trust
Differences-in-Differences Estimates?, 119 Q.J. ECON. 249 (2004). This problem also applies to
fixed-effects models. To make matters more complicated, others have argued that feasible general least
squares models produce more efficient estimators in some cases than those proposed by Bertrand and
colleagues. See Jerry Hausman & Guido Kuersteiner, Difference in Difference Meets Generalized Least
Squares: Higher Order Properties of Hypotheses Test (Boston Univ. Dep’t of Econ., Working Paper No.
WP2005-010, 2005), available at http://www.bu.edu/econ/workingpapers/papers/Guido/wp2005/
didgls23_MainB.pdf.
69. John Bian et al., Spillover Effects of State Mandated-Benefit Laws: The Case of Outpatient
Breast Cancer Surgery 4, 8–9, 18 tbl.1 (Sept. 12, 2007) (unpublished manuscript, on file with authors).
666
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
that general type is present or absent) or a series of variables that compares
different variations of the law to one another. The latter approach avoids the
dilution effects that may occur if some variants of the law are more effective
than others but all are lumped together in the analysis. But in some cases, as in
the mastectomy law example, not enough states have adopted laws of that type
to permit such detailed classification. Grouping heterogeneous legal mandates
together can lead to greater parsimony, but at the price of less helpful results.
These challenges require considerable expertise and resources to surmount.
For reasons we discuss later in the Article, scholars in law schools may struggle
to find them. In summary, the field of EHL presents a wealth of opportunities
for scholars to make important contributions to expanding the relatively small
knowledge base; however, these opportunities are costly to pursue. The EHL
literature to date has been dominated by descriptive analyses and quasiexperimental designs. The challenge for empirical legal researchers is to find
the proper balance between expediency and rigor70 in selecting a study design,
keeping in mind the quality of evidence they would like to contribute and the
constraints of the research environment in which their work is conducted. It is
also to ensure that their findings are interpreted—in their own reports and in
policy debates—in a way that reflects the limitations of their data and analytical
methods.
With this survey of the field in mind, we now turn to the question of how
EHL scholarship has impacted policy.
III. THE IMPACT OF EMPIRICAL HEALTH LAW SCHOLARSHIP ON POLICY:
TWO CASE STUDIES
In this Part, we juxtapose two policy histories that illustrate, respectively, the
promise and the pitfalls of building a reliable evidence base for legal reforms
and ensuring that policymakers respond to it. As an example of successful,
evidence-based health law, we offer the history of laws aimed at reducing motor
vehicle injuries to young drivers. As an illustration of where policymaking can
depart significantly from the best evidence about what is likely to succeed, we
offer the case of medical malpractice reform during the most recent insurance
“crisis.”
A. MOTOR VEHICLE SAFETY LAWS
Injuries from motor vehicle accidents are the leading cause of death among
youth aged sixteen to twenty in the United States.71 In 2000, there were 5600
70. Cf. Kaveh G. Shojania & Jeremy M. Grimshaw, Evidence-Based Quality Improvement: The
State of the Science, 24 HEALTH AFF. 138, 141–42 (2005) (describing a similar tradeoff in the choice of
designs for studying quality-improvement initiatives in health care).
71. NAT’L HIGHWAY TRAFFIC SAFETY ADMIN., U.S. DEP’T OF TRANSP., 2004 MOTOR VEHICLE OCCUPANT
PROTECTION FACTS 12 (2004), available at http://www.nhtsa.dot.gov/people/injury/airbags/MVOP2004/
images/2004MVOPFactsLo.pdf.
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
667
adolescent fatalities in traffic crashes, and the economic cost of accidents
involving drivers under age twenty-one was $32.8 billion.72 The high burden of
injury, coupled with the long history of state and federal regulation of highway
safety, has made vehicular injuries involving young people an obvious target for
legal interventions. These interventions have been plentiful (and not restricted
to adolescents). The best known are seatbelt laws, motorcycle helmet laws,
speed limits, laws relating to alcohol consumption and driving, and driver’s
education, testing, and licensing provisions.
Motor vehicle safety policy is an EHL research success story. The Centers for
Disease Control and Prevention has identified the massive reduction in motor
vehicle fatalities in the latter half of the twentieth century as one of the
century’s ten greatest public health achievements, and it has credited public
health law as a major contributor.73 Not only have public health laws in this
area proved effective, but evidence of their effectiveness has influenced their
spread across the states. There are many areas in which EHL research contributed to the widespread adoption of efficacious motor vehicle laws, but we will
focus on two of the best-documented successes: graduated driver licensing
(GDL) and minimum legal drinking age (MLDA) laws, both of which aim to
reduce traffic accidents involving young drivers.74
1. Graduated Driver Licensing
To address the risks arising from driver inexperience, GDL laws provide that
young drivers initially will be issued restricted licenses and will gain full
privileges only after they have acquired more experience. Typical GDL provisions require: holding of a learner’s permit for a minimum amount of time
(during which time the learner may only drive with an older person present);
restricting the number of passengers that young drivers may carry; prohibiting
freeway, highway, and nighttime driving; and requiring zero blood alcohol
content.75 In some countries, such as Canada, GDL schemes apply to all new
drivers regardless of age, but laws in the United States focus on young drivers.76
The story of GDL laws illustrates how empirical research can play an integral
72. Christine Branche et al., Graduated Licensing for Teens: Why Everybody’s Doing It, 30 J.L.
MED. & ETHICS 146, 146 (2002).
73. Ctrs. for Disease Control & Prevention, Ten Great Public Health Achievements—United States,
1900–1999, 281 JAMA 1481 (1999); Richard A. Goodman et al., Law and Public Health at CDC, 55
MORBIDITY & MORTALITY WKLY. REP. (SUPP.) 29 (2006) (describing the supportive role of laws and legal
tools).
74. We thank David Hemenway for suggesting these examples and referring us to pertinent research,
which he summarizes in his book, DAVID HEMENWAY, WHILE YOU WERE SLEEPING: SUCCESS STORIES IN
INJURY AND VIOLENCE PREVENTION (forthcoming 2008) (manuscript at 21, 23, on file with authors).
75. Herb M. Simpson, The Evolution and Effectiveness of Graduated Licensing, 34 J. SAFETY RES.
25, 27 (2003).
76. INS. INST. FOR HIGHWAY SAFETY & TRAFFIC INJURY RESEARCH FOUND., GRADUATED LICENSING: A
BLUEPRINT FOR NORTH AMERICA 1–2 (2004).
668
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
Figure 1. Lifecycle of an Empirical Health Law Success Story
role in both policy generation and policy evaluation, leading to successful,
evidence-based lawmaking. In brief, GDL laws emerged from public health
research highlighting a risk factor for serious injury. In response to these
findings, experimentation with GDL laws ensued in a small number of jurisdictions. The pioneer laws were evaluated rigorously and in a timely fashion. The
findings were impressive and led other jurisdictions to adopt similar laws. This,
in turn, led to additional evaluations and more widespread adoption of the laws.
This iterative process of research and policymaking constitutes an ideal “lifecycle” for a health law (Figure 1).77
The first stage in this lifecycle was the identification of a health risk factor
based on sound epidemiological research. In 1971, the North Carolina Highway
Commission reported findings indicating that young drivers were disproportionately represented in motor vehicle accidents.78 The risk was particularly elevated at night and when young drivers were accompanied by young passengers
in the front seat. These findings served as the original inspiration for GDL
77. This lifecycle bears some resemblance to the framework for evaluating complex interventions
for improving health suggested by Michelle Campbell and colleagues, which has four stages: “definition of components of the intervention,” development of study and intervention design, addressing
methodological issues, and “promoting effective implementation.” Michelle Campbell et al., Framework for Design and Evaluation of Complex Interventions to Improve Health, 321 BRITISH MED. J. 694
(2000).
78. Patricia F. Waller, The Genesis of GDL, 34 J. SAFETY RES. 17, 17–18 (2003).
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
669
proposals.79
The earliest proposals were made in the late 1970s by a transportation safety
researcher, and the National Highway Transportation Safety Administration
(NHTSA) developed a model GDL program in 1977.80 The idea did not
immediately take root among policymakers, however; only a few states implemented GDL legislation over the next decade.81 Meanwhile, other countries
were also considering GDL schemes. In 1983, an international symposium on
young drivers held in Canada endorsed the idea of GDL systems.82 The notion
received further support after publication in 1990 of a highly influential Canadian study that definitively linked increased driver experience with decreased
collision rates.83
The second stage of the lifecycle—pioneering policy experiments—emerged
around this time. In 1987, New Zealand implemented a GDL program in
response to data demonstrating that young drivers were at a highly elevated risk
for accidents and also in response to suggestions from a leading traffic safety
researcher that a GDL system would be helpful.84 This pioneering law was
studied from its earliest days in order to ascertain its effectiveness. Anecdotal
evidence of reductions in traffic accidents began to surface in New Zealand
almost immediately. A series of more rigorous evaluations, the first of which
was published in 1992, showed dramatic results.85
This third stage in the lifecycle—evaluation of policy experiments—led
swiftly to the fourth—emulation of successful experiments. Ontario, Canada
adopted a GDL program in 1994 based on the New Zealand results. Nova Scotia
did the same later that year, and four more provinces followed over the next five
years.86
These laws, too, were evaluated and shown to be highly effective. In 1998, a
79. Id. at 18.
80. U.S. DEP’T OF TRANSP., MODEL FOR PROVISIONAL (GRADUATED) LICENSING OF YOUNG NOVICE
DRIVERS (1977).
81. See Simpson, supra note 75, at 27. There was some limited state action in response to the
NHTSA report—Maryland adopted a nighttime driving restriction in 1979 and California and Oregon
introduced provisional licenses in 1983 and 1989, respectively—but comprehensive GDL schemes did
not emerge until the 1990s. Allan F. Williams, Earning a Driver’s License, 112 PUB. HEALTH REP. 453,
455 (1997).
82. YOUNG DRIVER ACCIDENTS: IN SEARCH OF SOLUTIONS (D.R. Mayhew et al. eds., 1983).
83. D.R. MAYHEW & H.M. SIMPSON, NEW TO THE ROAD: YOUNG DRIVERS AND NOVICE DRIVERS: SIMILAR
PROBLEMS AND SOLUTIONS? (1990).
84. Dorothy Begg & Shaun Stephenson, Graduated Driver Licensing: The New Zealand Experience,
34 J. SAFETY RES. 99, 99–100 (2003) (reporting that drivers aged fifteen to nineteen accounted for 27%
of all drivers in traffic accidents notwithstanding the fact that they drove only 8% of the total mileage
driven in New Zealand) (citing N.Z. MINISTRY OF TRANSP., A GRADUATED DRIVER LICENSING SYSTEM
(1985)).
85. Simpson reports that the initial evaluations of the New Zealand program found reductions in the
per-capita crash rate on the order of 25%, while later studies found longer-term effects around 7%.
Simpson, supra note 75, at 29; see also Begg & Stephenson, supra note 84, at 100–03 (summarizing
study findings).
86. Simpson, supra note 75, at 28.
670
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
pre-post analysis of the Ontario law showed a 31% decrease in the collision rate
for novice drivers, a 27% drop in alcohol-related collisions, a 62% drop in
collisions between midnight and 5:00 a.m., and a 61% decline in freeway
crashes.87 Controlled studies of the Canadian programs were available by the
year 2000, showing reductions in Nova Scotia’s per-capita crash rate in the 24%
to 36% range88 and a 5% reduction in fatalities and 14% reduction in injuries in
Quebec.89
The Canadian laws appear to have sparked interest in GDL in the United
States. In 1996, Florida became the first U.S. state to adopt GDL.90 By 1997, six
states had adopted GDL laws. Around this time, the Insurance Institute for
Highway Safety and the Traffic Injury Research Foundation of Canada issued a
joint report, which subsequently was updated several times, outlining detailed
recommendations for GDL programs.91 By 1999, seventeen states had adopted
them.
The early state laws were evaluated in relatively short order, and a series of
studies published around 2001 reported uniformly significant results.92 The
magnitude of the decreases in rates of fatal crashes, injury crashes, and all
crashes varied across states, but in some states the decrease was staggering:
North Carolina, for example, experienced a 57% drop in the per-capita fatal
crash rate among sixteen-year-old drivers.93 Based on data reported in a 2004
review of thirteen studies evaluating seven GDL schemes in the United States
and five in foreign countries, the median reduction in per-capita crashes was
31% among sixteen-year-olds and the median reduction in injury crashes was
28%.94 Statistically significant effects of the laws have been observed in all
states that have been studied.95 Many of these studies used quite sophisticated
designs to control not only for confounding variables but also for uncertainty
87. PAUL BOASE & LEO TASCA, ONT. MINISTRY OF TRANSP., GRADUATED LICENSING SYSTEM EVALUATION:
INTERIM REPORT (1998).
88. Daniel R. Mayhew et al., Impact of the Graduated Driver Licensing Program in Nova Scotia, 2
J. CRASH PREVENTION & INJURY CONTROL 179 (2001), cited in Simpson, supra note 75, at 29. Subsequent
studies confirmed this impact, but found that most of it occurred during the learner’s permit stage rather
than among drivers who held a restricted license. Simpson, supra note 75, at 29.
89. Simpson, supra note 75, at 30 (citing Joann Bouchard et al., The Quebec Graduated Licensing
System for Novice Drivers: A Two-Year Evaluation of the 1997 Reform, PROCEEDINGS OF THE 15TH
INTERNATIONAL CONFERENCE ON ALCOHOL, DRUGS, AND TRAFFIC SAFETY, STOCKHOLM, SWEDEN (2000)).
90. Simpson, supra note 75, at 28.
91. The most recent version is INS. INST. FOR HIGHWAY SAFETY & TRAFFIC INJURY RESEARCH FOUND.,
supra note 76.
92. See studies collected in Simpson, supra note 75, at 31 tbl.2. Another excellent literature review
is Jean T. Shope & Lisa J. Molnar, Graduated Driver Licensing in the United States: Evaluation Results
from the Early Programs, 34 J. SAFETY RES. 63 (2003).
93. Robert D. Foss et al., Initial Effects of Graduated Driver Licensing on 16-Year-Old Driver
Crashes in North Carolina, 286 JAMA 1588, 1588 (2001).
94. L. HARTLING ET AL., COCHRANE COLLABORATION, GRADUATED DRIVER LICENSING FOR REDUCING
MOTOR VEHICLE CRASHES AMONG YOUNG DRIVERS (REVIEW), 24 tbl.04, 25 tbl.06 (2006).
95. Shope & Molnar, supra note 92, at 67.
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
671
about the time frame in which the effects of the laws would be seen.96 Overall,
the evidence for the laws’ effectiveness is strong.97
Over the past decade, the results of the evaluations have led forty-seven states
and the District of Columbia to adopt some kind of GDL law.98 GDL researchers have noted that “the primary driving force” in the proliferation of GDL laws
across states has been not merely “the cascading effect of follow the leader” but
the scheme’s “proven effectiveness.”99 Thus, in GDL laws we see initial policy
experiments maturing into a dominant policy approach through an iterative
cycle of evaluation and emulation.100
2. Minimum Legal Drinking Age Laws
The MLDA is the best-studied alcohol policy in the United States,101 and
another example of a success story for health law and EHL research. Aimed at
reducing alcohol consumption by young adults, the MLDA has been set by state
law and historically has varied from state to state between eighteen and twentyone. In 1970, thirty-three states set the age at twenty-one. Between 1970 and
1975, twenty-nine states lowered their MLDA to eighteen, nineteen, or twenty.
This occurred as part of the broader social recognition of eighteen- to twenty-yearolds as legal adults that occurred with the passage of the Twenty-Sixth Amendment in response to the Vietnam War draft.102
When the MLDA was first lowered, it provoked considerable social controversy concerning the effect on alcohol-related motor vehicle accidents involving
young drivers. The available evidence at this point in time consisted largely of
small, uncontrolled studies, fueling debate about whether observed increases in
crash rates in states that lowered the MLDA were due to the legal change or
96. See, e.g., Michael R. Elliott & Jean T. Shope, Use of a Bayesian Changepoint Model To Estimate
Effects of a Graduated Driver’s Licensing Program, 1 J. DATA SCI. 43 (2003); John D. Langley et al.,
An Evaluation of the New Zealand Graduated Driver Licensing System, 28 ACCIDENT ANALYSIS &
PREVENTION 139 (1996); see also Hartling et al., supra note 94, at 6 (summarizing study designs in a
comprehensive literature review).
97. See HARTLING ET AL., supra note 94, at 9.
98. INS. INST. FOR HIGHWAY SAFETY, U.S. LICENSING SYSTEMS FOR YOUNG DRIVERS (2007), available at
http://www.iihs.org/laws/pdf/us_licensing_systems.pdf. This resource provides a detailed summary of
the features of each state’s GDL law.
99. Simpson, supra note 75, at 29.
100. In reviewing the literature on GDL laws, we also examined whether empirical legal research
was influential in judicial decision making, for example, in adjudicating constitutional challenges to the
laws. We found that GDL laws were not challenged in the courts.
101. See Alexander C. Wagenaar & Traci L. Toomey, Effects of Minimum Drinking Age Laws:
Review and Analyses of the Literature from 1960 to 2000, 14 J. STUD. ALCOHOL (SUPP.) 206, 206 (2002)
[hereinafter Wagenaar & Toomey, Effects] (citing Alexander C. Wagenaar & Traci L. Toomey, Alcohol
Policy: Gaps Between Legislative Action and Current Research, 27 CONTEMP. DRUG PROBS. 681 (2000));
see also id. at 206 (counting 132 studies on the effect of MLDA changes published between 1960 and
1999).
102. Alexander C. Wagenaar, Research Affects Public Policy: The Case of the Legal Drinking Age in
the United States, 88 ADDICTION (SUPP.) 75S, 76S (1993).
672
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
broader secular forces, such as improved police reporting practices.103
By the mid-1970s, controlled studies had become available. These generally
confirmed the initial study findings: a lower MLDA was significantly associated
with elevated risk of alcohol-related crashes among young drivers and increased
alcohol consumption.104 Effect sizes varied somewhat across studies, but clustered in the range of a ten to thirty percent increase in crashes.105
In light of this evidence, many states moved back in the other direction.
Maine was the first, in 1977, and several others quickly followed. The experience of these early movers was promptly studied, and the evaluations demonstrated a reduction in alcohol-related crashes. By 1983, sixteen states had
moved to a MLDA of twenty-one. Researchers continued to evaluate these
changes, and by 1983, five studies had been published on the effect of raising
the MLDA on traffic crashes; four demonstrated statistically significant effects
on at least one type of crash.106
Political interest groups, particularly Mothers Against Drunk Driving, used
these findings to press the case for reform at the national level.107 Their
advocacy was extremely effective, and in April 1982, President Reagan established a Presidential Commission on Drunk Driving at the urging of more than
300 members of Congress.108 While the Commission was carrying out its work,
the National Transportation Safety Board recommended a national minimum
drinking age of twenty-one.109 This recommendation appeared in the Commission’s interim report in December 1982 and headlined its final report in November 1983.110
The Commission’s work quickly led to congressional action. In October
1982, the Congress enacted the Alcohol Traffic Safety and National Driver
Register Act,111 which mandated that the Secretary of Transportation consider a
state MLDA of twenty-one as a criterion for awarding supplemental grants to
states. In December, it inserted a statement in its highway funding act that “[t]he
Congress strongly encourages each State to prohibit the sale of alcoholic
103. Id.
104. See id. at 207, 210 (listing studies on lowering the MLDA and their findings).
105. Id. at 209, 212–13, 215 (tallying outcomes among 132 studies).
106. See Wagenaar & Toomey, Effects, supra note 101, at 210.
107. Mary Pat Treuthart, Lowering the Bar: Rethinking Underage Drinking, 9 N.Y.U. J. LEGIS. &
PUB. POL’Y 303, 311–12 (2005–2006); Wagenaar, supra note 102, at 76S, 79S.
108. MIGDON SEGAL, CONG. RESEARCH SERV., DRUNK DRIVING AND THE NATIONAL DRIVER REGISTER 2
(1985).
109. Id. at 3.
110. PRESIDENTIAL COMM’N ON DRUNK DRIVING, FINAL REPORT: THE NEW HOPE OF SOLUTION 10–11
(1983).
111. Pub. L. No. 97-364, 96 Stat. 1738 (1982) (codified as amended at 23 U.S.C. § 408 (2000)); see
also National Highway Transportation Safety Administration, Incentive Grant Criteria for Alcohol
Traffic Safety Programs, 48 Fed. Reg. 5545, 5549 (Feb. 7, 1983) (announcing a final rule incorporating
the MLDA requirement and citing supportive research findings from the Presidential Commission and
the National Transportation Safety Board).
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
673
beverages to persons who are less than 21 years of age.”112
When the states did not immediately adopt laws enshrining this recommendation, Congress sought to increase its leverage. Initial legislation proposed a
national minimum drinking age to be imposed on states by Congress under its
interstate commerce power.113 The Reagan Administration had originally opposed a federal mandate, preferring to inspire states to act on their own, but
quickly reversed its position.114 The initial legislation nonetheless failed, but
Congress responded with an alternative proposal. In 1984, it passed the National
Minimum Drinking Age Act (NMDA),115 which, rather than imposing a national drinking age outright, tied the national minimum drinking age to federal
highway funds. The Act withheld five percent of highway funding from states
that did not raise their MLDA to twenty-one,116 increasing to ten percent
beginning in fiscal year 1988.117
Empirical legal research played prominently in hearings on both versions of
the NMDA legislation. Witnesses offering testimony cited research by University of Michigan researchers, the Insurance Institute for Highway Safety, the
National Highway Traffic Safety Administration, and law professors at the
University of Chicago.118 Although there was some dispute among witnesses
about how to interpret and credit the studies, the Congress ultimately appeared
to find the evidence persuasive. The NMDA was passed, and in response, all
states moved their MLDA to twenty-one by 1988.119
Research continued to be conducted after implementation of the NMDA, and
112. Surface Transportation Assistance Act of 1982, Pub. L. No. 97-424 § 209, 96 Stat. 2097 (1982).
113. A Bill to Restrict the Sales of Alcoholic Beverages in Interstate Commerce, H.R. 3870, 98th
Cong. (1983).
114. SEGAL, supra note 108, at 3.
115. Pub. L. No. 98-363, § 6(a), 98 Stat. 435, 437 (1984) (codified as amended at 23 U.S.C. § 158
(2000)).
116. Id. §6(a)(1).
117. 23 U.S.C. § 158(a)(1) (2000). The 1988 sunset provision on the original legislation was
removed by the Consolidated Omnibus Budget Reconciliation Act of 1985, which made the ten percent
withholding permanent. Pub. L. No. 99-272, § 4104a, 100 Stat. 82, 114 (1986).
118. National Minimum Drinking Age: Hearing Before the Subcomm. on Alcoholism and Drug
Abuse of the S. Comm. on Labor and Human Resources, 98th Cong. 38 (1984) (testimony of Christine
Lubinski, Director, Public Policy Office, National Council on Alcoholism); id. at 80–81 (testimony of
Michael M. Birkley, Executive Director, Tavern League of Wisconsin); id. at 130–31 (statement of the
National Restaurant Association); Prohibit the Sale of Alcoholic Beverages to Persons Under 21 Years
of Age: Hearing on H.R. 3870 Before the Subcomm. on Commerce, Transportation, and Tourism of the
H. Comm. on Energy and Commerce, 98th Cong. 10–11 (1983) (statement of Candy Lightner,
President, Mothers Against Drunk Drivers); id. at 59 (written testimony of Michael M. Birkley,
Executive Director, Tavern League of Wisconsin); id. at 66–67 (written testimony of Allan F. Williams,
Senior Behavioral Scientist, Insurance Institute for Highway Safety); id. at 73–74 (statement of Charles
A. Hurley, Executive Director for Federal Affairs, National Safety Council); id. at 240 (statement of
Ben Kelley, Senior Vice President, and Allan F. Williams, Senior Behavioral Scientist, Insurance
Institute for Highway Safety).
119. Wagenaar & Toomey, Effects, supra note 101, at 206.
674
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
these studies confirmed the earlier findings.120 A comprehensive evaluation of
the evidence base requested by Congress and reported in 1987 by the General
Accounting Office (GAO)121 raised the profile of the evidence and dispelled any
lingering doubts about the effectiveness of raising the drinking age. The GAO’s
thorough, 111-page review concluded that “the evidence is persuasive” that
raising the MLDA has significant effects on alcohol-related crashes among
eighteen- to twenty-year-olds, and that the observed effects were consistent
across studies in different states and with different designs and methods.122
The GAO scrutinized the methodologies of the existing studies and found
that although the studies varied in strength, a number of reliable, high-quality
analyses were available. The designs of MLDA studies have ranged from
simple pre/post studies to cross-sectional studies to the more tightly controlled,
longitudinal and time-series analyses. The best analyses123 exploit not only
variation over time as the law changed, but also variation across age groups
(eighteen- to twenty-year-olds, who were affected by the law, to twenty-one-andover groups, who were not).
MLDA laws have been challenged infrequently in court, but the three cases
that have been litigated have produced memorable results. The earliest case was
an equal-protection challenge to a Michigan constitutional amendment raising
the MLDA from nineteen to twenty-one. The federal district court’s opinion in
Felix v. Milliken124 contained an exhaustive discussion of two studies of Michigan’s law by Dr. Richard Lee Douglass on the effect of lowering the drinking
age on traffic accidents in Michigan.125 The court’s exposition of the studies is
remarkably detailed, and after this review of the evidence, the court had little
trouble in concluding that Michigan’s voters had a reasonable basis for changing the law.126
The second case was South Dakota’s famous challenge to the NMDA, which
120. See id. at 211–12 (collecting studies). Some of these studies have focused on the interactions of
the MLDA law change with other alcohol-related laws. See, e.g., William R. Ponicki et al., Joint
Impacts of Minimum Legal Drinking Age and Beer Taxes on US Youth Traffic Fatalities, 1975 to 2001,
31 ALCOHOLISM: CLINICAL & EXPERIMENTAL RES. 804 (2007) (examining interaction with beer taxes);
Robert B. Voas et al., Assessing the Effectiveness of Minimum Legal Drinking Age and Zero Tolerance
Laws in the United States, 35 ACCIDENT ANALYSIS & PREVENTION 579 (2003) (examining interaction with
0.02% blood alcohol concentration limit for drivers under age twenty-one). Twenty-four evaluations of
MLDA laws were published in the two years following the adoption of the NMDA alone. U.S. GEN.
ACCOUNTING OFFICE, DRINKING-AGE LAWS: AN EVALUATION SYNTHESIS OF THEIR IMPACT ON HIGHWAY
SAFETY 11 (1987).
121. U.S. GEN. ACCOUNTING OFFICE, supra note 120.
122. Id. at 39.
123. See, e.g., Alexander C. Wagenaar, Effects of an Increase in the Legal Minimum Drinking Age, 2
J. PUB. HEALTH POL’Y 206, 211–12 (1981); Alexander C. Wagenaar, Raising the Legal Drinking Age in
Maine: Impact on Traffic Accidents Among Young Drivers, 18 INT’L J. ADDICTIONS 365 (1983).
124. 463 F. Supp. 1360 (E.D. Mich. 1978).
125. Id. at 1379–82. Michigan lowered the drinking age to eighteen in 1972, and then raised it to
nineteen in 1978. Id. at 1364.
126. Id. at 1389.
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
675
culminated in the U.S. Supreme Court’s decision in South Dakota v. Dole,127
upholding the statute against challenges under the Spending Clause and the
Twenty-First Amendment. The Court’s Spending Clause analysis cited findings
of the Presidential Commission, which had concluded that varying drinking
ages across the states fostered drunk driving by encouraging young people to
commute to states where the drinking age was lower,128 and concluded that the
Act pursued “the general welfare”129 and directly related to the federal purpose
for which highway funds were spent.
The third case, Manuel v. Louisiana,130 is a relatively recent challenge to a
post-NMDA state law. Ruling on an equal-protection challenge, the Louisiana
Supreme Court held that the MLDA survived even intermediate scrutiny due to
the extensive evidentiary record that the higher MLDA substantially furthered
the state’s important interest in reducing alcohol-related crashes involving
youth.131 The court’s decision on rehearing overturned its initial holding132 on
the grounds that it had been based on the wrong piece of statistical evidence.133
The rehearing opinion contained in an appendix an extensive discussion of the
evidence base on MLDA laws,134 specifically referencing the GAO’s report as
proof that the efficacy of MLDA laws was “widely accepted by experts.”135 In
summary, the courts, as well as state and federal legislatures, have made good
use of high-quality empirical research on the efficacy of MLDA laws.
Overall, MLDA laws have had a lifecycle similar to GDL laws, characterized
by iterative interactions between lawmaking and empirical research. The MLDA
story is even more dynamic than the history of GDL laws, however, because of
the presence of not one, but two, natural experiments. Twenty-nine states first
lowered their MLDA laws, then sixteen states reversed that decision within a
short period of time. After the drinking age was lowered, empirical evaluations
provided evidence that the change in the law created a health risk. A commonsense solution was clear, and many states adopted it. These changes in the law
were then extensively evaluated and the positive findings disseminated to state
and federal policymakers through the efforts of political interest groups. They
were deemed sufficiently compelling to support federal action, which led to
universal adoption by the states. Finally, research and evaluation continued well
127. 483 U.S. 203 (1987).
128. Id. at 209 (citing PRESIDENTIAL COMM’N ON DRUNK DRIVING, supra note 110, at 11).
129. Id. at 206.
130. 677 So. 2d 116 (La. 1996).
131. Id. at 125.
132. Manuel v. Louisiana, 692 So. 2d 320, 321, rev’d, 677 So. 2d 116, 118 (La. 1996).
133. The court originally held that the state had not demonstrated the efficacy of its scheme because
the number of alcohol-related crashes was higher among drivers aged twenty-one to twenty-three than
among those aged eighteen to twenty. Manuel, 692 So. 2d at 327. On rehearing, the court decided that
the correct statistic was the proportion of alcohol-related traffic crashes attributable to each group.
Finding that the younger group was proportionally the most dangerous group, it upheld the law.
Manuel, 677 So. 2d at 122–23.
134. Manuel, 677 So. 2d at 128–32.
135. Id. at 125.
676
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
after universal adoption, providing a continually growing evidence base that can
be used to head off efforts to reverse the laws or challenge them on constitutional grounds.
This is an unqualified success story for EHL research. The National Highway
Traffic Safety Administration has estimated that 17,359 traffic fatalities were
averted between 1975 and 1997 as a result of a higher MLDA.136
3. Enabling and Inhibiting Factors
What accounts for the success of EHL research in the area of motor vehicle
safety? Four enabling factors appear to have considerable explanatory force: the
presence of natural experiments, the ready availability of data, the uncontested
framing of the issue as a public health problem, and the facilitating role of
political interest groups.
First, natural experiments with changes to GDL and MLDA laws provided
opportunities for controlled testing of hypotheses about the efficacy of these
changes. This was particularly true for MLDA laws because the pioneering
experiments took place in the United States and because states moved their
MLDA from high to low and back again, providing two points at which to
observe the effects of this change.
Second, researchers were able to take advantage of these natural experiments
because of the ready availability of good data on the outcome variables of
interest. The primary outcome variable for GDL laws is the overall rate of
motor vehicle crashes involving young drivers. Secondary outcomes include
rates of injury crashes, fatal crashes, hospitalizations related to crashes, crashes
during nighttime, alcohol-related crashes, and crashes involving property damage.137 In MLDA studies, the outcome variables have included traffic crashes
involving alcohol, rates of alcohol consumption among under-twenty-one-yearolds, alcohol sales levels, and rates of driving while under the influence of
alcohol.138
For some of these measures, researchers have had to rely on self-reports on
youth surveys, impressions of police officers responding to accidents, or statistical surrogates for alcohol involvement in a crash (for example, crashes that
occur at nighttime and involve an unaccompanied male driver).139 In general,
however, the quality of the data has been good, and its availability has been
very good, compared to other areas of health law. State departments of public
safety, police departments, or divisions of motor vehicles maintain electronic
databases of motor vehicle accidents. Characteristics of drivers are often available from police reports or by searching driver’s license databases. Reliable
136. NAT’L HIGHWAY TRAFFIC SAFETY ADMIN., TRAFFIC SAFETY FACTS 1997: ALCOHOL 4 (1998).
137. HARTLING ET AL., supra note 94, at 3.
138. Wagenaar & Toomey, Effects, supra note 101, at 207–12 (listing outcome measures in all
MLDA studies published between 1960 and 1999); U.S. GEN. ACCOUNTING OFFICE, supra note 120, at
21–23.
139. U.S. GEN. ACCOUNTING OFFICE, supra note 120, at 24–25.
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
677
measures of alcohol involvement in crashes—for example, breath tests and
blood-alcohol content tests conducted by police officers and coroners—have
often been available.
Since the 1980s, the federal government and many states have invested in
maintaining and improving their traffic crash databases. This has enabled
researchers to overcome some of the criticisms of early studies—for example,
the objection that many MLDA studies relied on proxy measures of alcohol
involvement in crashes rather than actual evidence of driver impairment.140
Advances in statistical modeling and computing in the 1980s and 1990s also
made it possible for researchers to employ more sophisticated models that
addressed concerns about confounds in studies from the 1970s and took full
advantage of the available data.
A third enabling factor had to do with the framing of the issue in policy
debates. From the earliest days of the debates, motor vehicle accidents involving young people were framed as a public health problem. Research had clearly
established the high rate of crashes involving young drivers and the devastating
injuries that resulted. Injury prevention was an established subfield of public
health, and prevention of harm to children was a core concern of the field. As a
result, it was natural to look to public health research for policy solutions.
Indeed, framing the issue as a public health problem attracted the interest of a
number of skilled public health researchers. The overwhelming bulk of the
empirical scholarship on GDL and MLDA laws was conducted not by legal
scholars, but by researchers in the public health community. Finally, the clear
public health ramifications of the problem drew the interest of influential and
relatively apolitical federal agencies, such as the NHTSA, which provided
funding and a receptive ear for empirical research.
The fourth enabling factor was the facilitative role played by political interest
groups. In the case of GDL laws, no influential interest groups emerged to
oppose the policy change or dispute the evidence base that supported it. This
was partially related to the issue’s framing; characterizing the problem as a
public health threat and a pediatric problem made it a highly sympathetic
cause.141 No interest group contested that reducing the rates of crashes due to
driver inexperience or alcohol use among young people was a worthy target for
140. Id. at 13.
141. In contrast, motorcycle helmet laws, which have also been proven to be highly effective in
preventing injuries, have faced strong opposition on antipaternalism grounds, probably because they
affect primarily adults. See, e.g., City of Adrian v. Poucher, 247 N.W.2d 798 (Mich. 1976) (upholding
Michigan’s motorcycle helmet law despite defendant’s claim that the law unlawfully invades individual
rights). Indeed, the recent trend in state motorcyle helmet laws has been to scale them back so that they
cover only young people. Marian Moser Jones & Ronald Bayer, Paternalism and Its Discontents:
Motorcycle Helmet Laws, Libertarian Values, and Public Health, 97 AM. J. PUB. HEALTH 208, 215
(2007) (“In 1997, after pressure from state-level motorcycle activists, Arkansas and Texas repealed their
universal helmet laws and instead required helmets only for riders aged younger than 21 years. These
repeals were followed by similar actions in Kentucky (1998), Louisiana (1999), Florida (2000), and
Pennsylvania (2003).”).
678
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
policy interventions. This paved the way for rapid policy responses. The only
industry group that played prominently in the policy debate was the Insurance
Institute for Highway Safety, which favored GDL laws and funded and participated in studies to build the evidence base for them.142 In the case of MLDA
laws, there was organized opposition to raising the drinking age, primarily
alcohol industry groups. But interestingly, the influence of these organizations
may actually have helped promote high-quality research. They did not produce
a raft of poor-quality studies of their own,143 but rather dedicated themselves to
criticism of academic research demonstrating a public health benefit to raising
the MLDA. Researchers were aware that these groups would try hard to
discredit their studies and, as a result, took a great deal of care in designing
them, so it would be difficult to find scientific flaws.144 Political interest groups
that favored raising the MLDA were also crucial to the success story of MLDA
laws. Groups such as Mothers Against Drunk Driving partnered with researchers who were committed to ensuring that their findings reached a policy
audience, resulting in a highly successful dissemination effort.145
To be sure, these enabling factors operated in the face of several factors that
inhibited the successful production of reliable empirical legal research. These
were primarily methodological problems. A key shortcoming of the empirical
base regarding GDL laws is the studies’ consistent focus on evaluating GDL
laws one state (or country) at a time. Single-state studies suffer from clear
generalizability issues. Perhaps more importantly, because each state’s GDL
scheme consists of a number of different features, finding that one state’s
scheme had a bigger effect than another state’s provides little information about
exactly which features were most important.146 GDL laws are highly variable
across states—for example, in the age groups covered and the particular restrictions imposed.147 This heterogeneity may help explain the variability in the
magnitude of effects found among studies of GDL laws in different states.148
Along with variations in study methodologies, it makes it difficult to directly
compare individual states’ evaluation results.149
For MLDA, one challenge for empirical evaluation of the laws has been that
studies have found very low rates of enforcement actions (such as compliance
142. Cf. Simpson, supra note 75, at 28 (noting that the Institute “had played a dominant leadership
role for years, not only in the research that provided the groundwork for graduated licensing but also in
actively promoting its adoption”).
143. In other debates, industry groups have produced their own studies to counter those produced by
organizations advocating changes disfavoring them. See, e.g., infra notes 180–82 and accompanying
text.
144. Wagenaar, supra note 102, at 80S.
145. Id.
146. Shope & Molnar, supra note 92, at 67 (“Understanding which components of GDL have
contributed to the crash reductions is difficult because programs are implemented as a total package.”).
147. See INS. INST. FOR HIGHWAY SAFETY, supra note 98 (outlining several dimensions of variation).
148. Simpson, supra note 75, at 33.
149. See Shope & Molnar, supra note 92, at 64.
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
679
checks on alcohol retailers) in some states.150 If alcohol sellers are aware that
the probability of an enforcement action is low, they may not comply with
MLDA laws. Assuming that this is likely the case, effect sizes observed in
MLDA studies probably underestimate the potential effects of raising the
MLDA on traffic crashes were states to enforce the laws vigorously.
Despite these methodological issues, the empirical research base around
MLDA and GDL laws is impressive, and policymakers’ response has been
equally impressive. Motor vehicle safety laws are one area that demonstrates
the power of empirical legal research to contribute to enormous reductions in
the burden of injuries.
B. MEDICAL MALPRACTICE REFORM
1. The Production of Scholarship and Policy
In striking contrast to the story of motor vehicle safety laws, the story of
medical malpractice reform laws exemplifies a missed opportunity to use
rigorous EHL research to develop evidence-based policy. Legislative consideration of liability-limiting tort reforms for medical malpractice has a long and
cyclical history151—interest in reform waxes and wanes with trends in the cost
of medical malpractice liability insurance. Despite the accretion of a substantial
base of scholarly research studying the effects of these reforms over time,152 the
policy debates have had the same disappointing tone, focus, and outcomes over
three successive malpractice “crises,” which are characterized by spikes in
malpractice insurance premiums and/or reductions in the supply of malpractice
insurance.
During the first malpractice crisis, in the mid-1970s, the empirical research
base was extremely thin. Little was known about the extent to which malpractice reforms achieved their four primary aims: stabilizing liability insurance
premiums, controlling the frequency of malpractice claiming, reducing the
severity (average cost) of malpractice claims, and preventing physician behaviors that are believed to be correlated with high-risk malpractice environments,
such as leaving practice and practicing defensive medicine.
The first serious attempt to synthesize the literature to aid policymaking came
in 1973 with the report of the Secretary of Health, Education and Welfare’s
Commission on Medical Malpractice,153 in which most of the studies identified
were not of sufficient rigor or scope to generate useful broad conclusions about
150. Wagenaar & Toomey, Effects, supra note 101, at 218.
151. Rogan Kersh, Medical Malpractice and the New Politics of Health Care, in MEDICAL MALPRACTICE AND THE U.S. HEALTH CARE SYSTEM 43, 43–67 (William M. Sage & Rogan Kersh eds., 2006)
[hereinafter MEDICAL MALPRACTICE IN THE U.S.], provides an excellent summary of this history.
152. To limit the scope of our discussion, we will focus on the role that empirical studies of
malpractice reforms have played during the most recent crisis from 2000 to 2006.
153. U.S. DEP’T OF HEALTH, EDUC. & WELFARE, REPORT OF THE SECRETARY’S COMMISSION ON MEDICAL
MALPRACTICE (1973).
680
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
the malpractice environment.154 In the late 1970s, the National Association of
Insurance Commissioners advanced the empirical base significantly by collecting and tabulating three years’ worth of insurance-carrier data on closed malpractice claims.155 The availability of this database made malpractice a more
attractive and feasible topic for empirical scholars to study.156
Although Patricia Danzon and other analysts contributed a handful of carefully conducted, controlled analyses of the effects of tort reforms over the next
decade,157 when the next malpractice crisis struck in the mid-1980s, policymakers still had relatively little in the way of empirical research to illuminate the
path to effective reforms.158 The federal government again appointed a task
force, which issued its findings as the crisis began to wane.159 Notwithstanding
this contribution, the policy debates tended, as in the 1970s, to be dominated by
political interest groups representing the medical profession on the one hand,
and the legal profession on the other.160 Scholars and their research findings
rarely played a role.161 The results were similar to those of the mid-1970s:
through the end of the 1980s and into the early 1990s, legislatures reached for a
limited set of reforms, such as damages caps and collateral-source offsets,
making few innovations.162 Notably, the few studies of damages caps available
at this time had determined that although they reduced the average value of
awards, they did not significantly affect the frequency of claiming or the level
of malpractice insurance premiums.163
Over the next fifteen years, the body of scholarly literature evaluating the
effects of tort reforms grew significantly. A number of top-flight economists,
154. Stephen Zuckerman et al., Information on Malpractice: A Review of Empirical Research on
Major Policy Issues, 49 LAW & CONTEMP. PROBS. 85, 86–87 (1986).
155. NAT’L ASS’N OF INS. COMM’RS, MALPRACTICE CLAIMS: FINAL COMPILATION (M. Patricia Sowka ed.,
1980).
156. Randall R. Bovbjerg & Joel M. Schumm, Judicial Policy and Quantitative Research: Indiana’s
Statute of Limitations for Medical Practitioners, 31 IND. L. REV. 1051, 1055 (1998).
157. Patricia Danzon, The Frequency and Severity of Medical Malpractice Claims, 27 J.L. & ECON.
115 (1984); Patricia M. Danzon, The Frequency and Severity of Medical Malpractice Claims: New
Evidence, 49 LAW & CONTEMP. PROBS. 57 (1986); Frank A. Sloan, State Responses to the Malpractice
Insurance “Crisis” of the 1970s: An Empirical Assessment, 9 J. HEALTH POL. POL’Y & L. 629 (1985).
158. See Zuckerman et al., supra note 154, at 110 (reporting that in 1985, “comparatively little
generalizable empirical data, and even less explanation or analysis of the causes or effects of such
quantitative phenomena, is available”).
159. U.S. DEP’T OF HEALTH & HUMAN SERVS., REPORT OF THE TASK FORCE ON MEDICAL LIABILITY AND
MALPRACTICE (1987).
160. See Bovbjerg & Schumm, supra note 156, at 1055–56.
161. Zuckerman et al., supra note 154, at 111.
162. Bovbjerg & Schumm, supra note 156, at 1055–56; Eleanor D. Kinney, Malpractice Reform in
the 1990s: Past Disappointments, Future Success?, 20 J. HEALTH POL. & POL’Y L. 99, 112 (1995).
Bovbjerg and Schumm note that one area of innovation was the creation of administrative compensation funds for serious neurological birth injuries in Virginia and Florida. Bovbjerg & Schumm, supra
note 156, at 1056.
163. These studies are summarized in Michelle M. Mello & David M. Studdert, Understanding
Medical Malpractice Damages Caps 30–43, 63–64 (Dec. 7, 2005) (unpublished manuscript, on file with
authors).
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
681
law professors, and public-policy scholars contributed studies that controlled
(more or less successfully) for variations in state environments that might
explain gaps in premium levels, physician supply, or claiming levels and
costs.164 When the third malpractice crisis hit, around the turn of the twenty-first
century, scholarly interest in the topic redoubled. Various arms of the federal
government published studies of varying quality in the early years of this crisis.
The lifecycle of academic studies was considerably longer, but by 2004, scholarly studies capturing data through the early 2000s began to surface.
The academic studies of this period have been, as a group, quite rigorous and
carefully conceived. They have made it possible to draw a few fairly definitive
conclusions about damages caps and other conventional tort reforms.165 Noneconomic damages caps appear to have statistically significant effects on premiums
and physician supply, but the effect sizes are quite modest.166 Caps have a larger
effect on award amounts, as might be expected, but there is no evidence that
they affect claims frequency.167 Joint-and-several liability reform has been
found to be significantly negatively associated with premium levels but not
other outcome variables.168 “Study findings [regarding shorter] statutes of
limitations [and] repose are mixed, but some strong studies have found an effect
on claims frequency and premiums . . . .”169 The weight of the evidence suggests that attorney-fee limits, collateral-source rule reform, pretrial screening
panels, and periodic payment have little to no effect on relevant outcome
variables.170
High-quality empirical research on malpractice reforms continues to be
produced at a high rate. The question is: is anyone listening? Unfortunately, the
answer for legislators appears to be “no.” Although the evidence taken as a
whole does not support a clear regulatory solution, broad conclusions about the
connection between malpractice reforms and outcome variables could be helpful for policymakers as they attempt to fashion sensible reforms. Despite this,
little of this research has found its way into policy debates over reforms during
the latest malpractice crisis. To the extent that it has, policymakers have tended
to use the studies more to support their ex ante policy preferences than to
determine optimal regulation.
The Congress has been focused narrowly on legislation proposing a federal
cap on noneconomic damages. The House and Senate held several hearings on
medical liability from 2003 through 2006 to consider this legislation (and, on a
164. For a full discussion, see id. at 30–47.
165. See MICHELLE M. MELLO, THE ROBERT WOOD JOHNSON FOUND., MEDICAL MALPRACTICE: IMPACT OF
THE CRISIS AND EFFECT OF STATE TORT REFORMS 9–12 (2006), available at http://www.rwjf.org/pr/synthesis/
reports_and_briefs/pdf/no10_researchreport.pdf.
166. Id. at 11–12.
167. Id. at 11.
168. Id. at 9.
169. Id.
170. Id.
682
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
smaller number of occasions, bills to authorize grants to support demonstration
projects of experimental tort reforms such as administrative compensation
programs and early-offer programs). The hearings on caps were remarkably
consistent in the composition of the panel of invited witnesses, the kinds of
arguments and evidence offered by the witnesses, and the tenor of the dialogue
between witnesses and committee members.171 The panels usually consisted of
one or more malpractice “victims” who told their personal stories, one or more
representatives of physician or hospital organizations, an insurance industry
representative, a representative of the trial bar, and a representative of a
consumer organization (often one with ties to the bar). Occasionally, a lone
academic would appear.
Many witnesses did cite data in their testimony, but they almost uniformly
relied on descriptive studies, eschewing more complex, controlled studies produced by academics.172 Testimony was liberally peppered with anecdotes about
persons affected by medical errors or high insurance costs.173 Questioning of
the witnesses by the committee members rarely addressed data, and when it did,
it typically focused on analyses conducted by interest groups rather than
academics.174
The policy process at the state level was somewhat better, in that a number of
state governors and legislators appointed study commissions to investigate the
evidentiary basis for a range of proposed reforms. These commissions, however,
tended to fall back on familiar strategies of information-gathering—namely,
assembling a panel of witnesses, most of which represented the various stakeholder groups. Little independent investigation and review of scholarly literature appears to have been conducted.
The experience of the Florida Governor’s Select Task Force on Healthcare
171. The hearings on demonstration projects featured a much larger role for academics and
discussion of scholarly studies. See generally Medical Liability: New Ideas for Making the System Work
Better for Patients: Hearing on S. 1337 Before the Comm. on Health, Education, Labor, and Pensions,
109th Cong. (2006); Innovative Solutions to Medical Liability: Hearing Before the Subcomm. on Health
of the H. Comm. on Energy and Commerce, 109th Cong. (2006).
172. See, e.g., Medical Liability Reform: Stopping the Skyrocketing Price of Health Care: Hearing
Before the H. Comm. on Small Business, 109th Cong. 13–14 (2005) (statement of Lawrence E. Smarr,
President, Physician Insurers Association of America) (citing simple descriptive statistics from studies
by the GAO, the National Association of Insurance Commissioners, and AM Best). But see Assessing
the Need to Enact Medical Liability Reform: Hearing on H.R. 5 Before the H. Comm. on Energy and
Commerce, 108th Cong. 42–48 (2003) (testimony of Fred J. Hiestand, President, Californians Allied for
Patient Protection) (citing several controlled studies along with simple descriptive studies).
173. See, e.g., Harming Patient Access to Care: The Impact of Excess Litigation: Hearing Before H.
Comm. on Energy and Commerce, 107th Cong. (2002) (testimony of Lauren Townsend, Director,
Coalition for Consumer Justice).
174. One colorful example is an exchange between Rep. John Conyers and an insurance association
executive, in which Conyers administered a “true or false” exam in an attempt to force the executive to
acknowledge the findings of a consumer group, the Center for Justice and Democracy, whose work the
executive believed to be of poor quality. See Help Efficient, Accessible, Low-Cost, Timely Healthcare
(HEALTH) Act of 2003: Hearing on H.R. 5 Before H. Comm. on the Judiciary, 108th Cong. 86–87
(2003).
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
683
Professional Liability Insurance is illustrative. Appointed by Governor Jeb Bush
in 2002, the Task Force was directed to examine Florida’s medical liability
environment and make recommendations for improving it. The Task Force had
a blue-ribbon roster, including former Secretary of Health and Human Services
Donna Shalala, and ended up recommending to the legislature that it adopt a
package of tort reforms. The proposed package included a $250,000 noneconomic damages cap, expert-witness qualifications, periodic payment, joint-andseveral liability reform, and small-scale demonstration projects of administrative
compensation.175 It took its work seriously, meeting ten times over five months,
holding several hearings, producing a 345-page report, and submitting thirteen
volumes of supporting materials along with it.176 Despite this seemingly comprehensive investigation and the fact that scholarly studies have generated evidence
in support of many of the reforms the Task Force recommended, nothing
indicates that the Task Force considered the academic research. Its discussion of
expert-witness qualifications, joint-and-several liability reform, and periodic
payment contained no citations to empirical studies of any kind and no references to information provided by academics.177 Its findings concerning administrative compensation also cited no scholarly studies of such systems, but did at
least reference a report by the Institute of Medicine and testimony from a
handful of scholars familiar with the literature.178 Its discussion of damages
caps is, from an academic’s perspective, highly unsatisfactory. The Task Force
referenced not a single published study of the effects of damages caps, relying
instead on proprietary analyses performed by a consultant, William Hamm,
under a contract with an advocacy organization whose purpose is to preserve
California’s cap on noneconomic damages.179
Similar processes and conclusions played out in other states. The dominance
of uncontrolled, descriptive studies conducted by political interest groups in
these debates has been striking. A typical report compares premiums in states
with caps on noneconomic damages, to premiums in other states that do not
have caps, without controlling for any of the ways in which the states may
differ.180 These studies suffered from numerous other methodological problems
including inappropriate comparison groups, failure to adjust for inflation, and
175. The Task Force’s work is described in its final report. GOVERNOR ’S SELECT TASK FORCE ON
HEALTHCARE PROF ’L LIAB. INS., FINAL REPORT vii–xvi (2003), available at http://www.doh.state.fl.us/
myflorida/DOH-Large-Final%20Book.pdf (listing the Task Force’s sixty recommendations).
176. Id.
177. See id. at 235–39, 260–61, 276–77.
178. Id. at 159–70.
179. See id. at 193–200. The organization, Californians Allied for Patient Protection, commissioned
a series of such reports from Dr. Hamm. Based on a review of the one report we could obtain from the
organization, the estimates generated by Dr. Hamm were not on very solid evidentiary footing (details
available from authors upon request); see WILLIAM G. HAMM, LECG, HOW THE MICRA CAP INFLUENCES
HEALTH CARE COSTS FOR SAFETY NET PROVIDERS AND MEDI-CAL. (1999).
180. See, e.g., Am. Ass’n for Justice, Medical Malpractice Insurance: Do Caps Reduce Insurance
Premiums?, http://www.atla.org/pressroom/FACTS/medmal/AvgPremiums_1104.aspx (last visited Oct.
22, 2007).
684
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
selective reporting of data.181 They were, however, presented simply, in straightforward language, with vivid charts and graphs, in short briefing papers that
were far more accessible (physically and cognitively) to policymakers than
scholarly journal articles.182 The academic articles described complicated multivariate regression models in lengthy and pointy-headed presentations—hardly
fodder for headline-grabbing hearings.
As a result, to the extent that enacted policies reflected empirical findings, it
was largely a matter of luck rather than deliberate process. Examining the
relative prevalence of various kinds of tort-reform legislation at the state level
over the past two years (Table 1),183 the most encouraging observation is that
the reform with the greatest amount of empirical support from well-designed
studies, damages caps, has been considered by more states than any other
reform. On the other hand, many other reforms with little or no evidence to
support them have been repeatedly considered and, on several occasions, enacted. Similar observations can be made from an examination of the prevalence
of state reforms enacted in recent years (Figure 2).184
At the federal level, the tort-reform debate during the last malpractice crisis
has been “much sound and fury, yielding no substantive policy change.”185 The
House repeatedly passed the federal damages cap legislation, and the Senate
repeatedly rejected it, until the Democrats took control of Congress following
the 2006 election and definitively killed the initiative. Overall, in assessing the
results of the federal reform effort, Rogan Kersh has written that “the contrast
between expectations and legislative outcome has never been starker in U.S.
malpractice politics.”186 A sunny reading of this history is that a litany of
single-minded lobbyists and second-rate analyses failed to sway Congress. A
realist view is that the advocates, and the legislators sympathetic to them,
quickly came to irremediable gridlock.
What about judicial decision making? This scholarship certainly is relevant to
questions of whether tort reforms reasonably advance a state interest, as re-
181. For a full discussion, see MELLO, supra note 165, at 22–23, and Mello & Studdert, supra note
163, at 20–30.
182. See WILLIAM HALTOM & MICHAEL MCCANN, DISTORTING THE LAW: POLITICS, MEDIA, AND THE
LITIGATION CRISIS 101, 102 (2004).
183. These frequencies were compiled by the National Conference of State Legislatures. See Nat’l
Conference of State Legislatures, Medical Malpractice Tort Reform: 2005 Enacted Legislation in the
States, http://www.ncsl.org/standcomm/sclaw/medmalenacted2005.htm (last visited Oct. 22, 2007);
Nat’l Conference of State Legislatures, Medical Malpractice Tort Reform: 2006 State Introduced
Legislation, http://www.ncsl.org/standcomm/sclaw/medmalreform06.htm (last visited Oct. 22, 2007);
Nat’l Conference of State Legislatures, State Medical Malpractice Reform: 2005 Numbers at a Glance,
http://www.ncsl.org/standcomm/sclaw/medmalataglance.htm (last visited Oct. 22, 2007).
184. Data come from Avraham, supra note 52, at 8. Year-to-year variations represent both new
adoptions of laws and existing laws being struck down by courts.
185. Kersh, supra note 151, at 49.
186. Id.
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
685
Table 1. Number of State Legislatures Considering and Enacting
Malpractice Reform Legislation, 2005–2006
2005
2006
Evidence of
Efficacy*
Considered Enacted Considered Enacted
Damages Caps
Some
28
5
14
1
Shortening
Statute of
Limitations
Some
13
1
5
1
Joint-andSeveral
Liability
Reform
Some
12
2
2
1
CollateralSource Rule
Reform
Virtually
None
9
0
5
0
Screening Panels
Virtually
None
13
2
5
0
Periodic Payment
None
11
0
4
0
Attorney-Fee
Limits
None
12
1
4
1
Expert Witness
Standards
Not Studied
17
5
7
0
* As measured by effects on insurance premiums, claim frequencies, and claims
costs.
quired by equal protection and due process analysis.187 Yet, empirical legal
studies are rarely discussed in these judicial opinions. For example, in rejecting
an equal protection challenge to Louisiana’s $500,000 damages cap in 1992, the
Louisiana Supreme Court considered no empirical evidence of the cap’s effectiveness whatsoever. Applying rational basis review, it simply concluded, without
marshaling facts or argument, that the legislation “furthers the state’s purpose of
compensating victims” by increasing the likelihood that physicians would be
187. See Ronen Avraham, An Empirical Study of the Impact of Tort Reforms on Medical Malpractice Settlement Payments 3 (Nw. Law Sch. Law & Econ. Research Paper Series, Paper No. 06-07,
2006), available at http://ssrn.com/abstract⫽912922 (finding little evidence of an association between
reforms and medical malpractice insurance premiums and arguing that courts should consider the
empirical evidence when evaluating the constitutionality of reforms). Most states require only a
showing of a reasonable relationship, but a few apply intermediate scrutiny and require that tort reforms
substantially advance the state’s interest. See Carly N. Kelly & Michelle M. Mello, Are Medical
Malpractice Damages Caps Constitutional? An Overview of State Litigation, 33 J.L. MED. & ETHICS
515, 522–23 (2005).
686
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
Figure 2. Prevalence of State Malpractice Reforms, 2000–2005
able to maintain insurance.188 A dissenting justice, in contrast, criticized the
majority for overlooking several empirical studies showing that caps had not
had their intended effects.189
An exception to the general trend is the Wisconsin Supreme Court’s 2005
decision in Ferdon v. Wisconsin Patients Compensation Fund,190 holding that
the state’s noneconomic damages cap violated equal protection.191 The court
cited several descriptive studies of damages caps that suggested caps were
unlikely to have much of an effect because only a small proportion of all awards
were high enough to implicate the cap. The court also discussed the limitations
of other studies investigating the connection between caps and malpractice
insurance premiums.192 It also cited descriptive studies concerning caps’ effects
on physician migration.193 The court’s discussion of these studies is relatively
sophisticated, compared to other judicial opinions, but that sophistication is
undermined by the fact that in selecting studies for citation, the court excluded
so much wheat and included so much chaff.194 Scrutiny of the citations reveals
188. Butler v. Flint Goodrich Hosp., 607 So. 2d 517, 521 (La. 1992).
189. Id. at 527 (Dennis, J., dissenting).
190. 701 N.W.2d 440 (Wis. 2005).
191. Id. at 491.
192. Id. at 469–71.
193. Id. at 485–86.
194. For example, the court relied heavily on two student papers and cited studies commissioned by
interest groups. See id. at 448, 470, 473, 474, 483, 484 (citing HEALTH INS. ASS’N OF AM., ISSUE BRIEF:
WHY DO HEALTH INSURANCE PREMIUMS RISE (2002), available at http://downloads.heartland.org/
14402.pdf; Kevin J. Gfell, Note, The Constitutional and Economic Implications of a National Cap on
Non-economic Damages in Medical Malpractice Actions, 37 IND. L. REV. 773 (2004); Elizabeth Stewart
Poisson, Comment, Addressing the Impropriety of Statutory Caps on Pain and Suffering Awards in the
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
687
the key selection criterion for studies: the court cited what it could read full-text
on the internet or in Westlaw or LexisNexis. Other works in medical and health
policy journals do not appear.
2. Enabling and Inhibiting Factors
When we look broadly at this history, several factors appear to have encouraged the production of empirical malpractice scholarship and held promise for
its potential incorporation into policymaking, but these factors have been
overshadowed by a number of substantial barriers.
On the research side, a key catalyst for growth in the field was the development of closed-claim datasets, such as the National Practitioner Data Bank and
state insurance department databases. This spurred a cottage industry in malpractice research among a cadre of well-known economists. The high visibility and
policy salience of the issue may also have helped attract highly qualified
scholars to the field.
Second, as with motor vehicle safety laws, malpractice reforms offer a large
number of natural experiments that create ready opportunities for study. The
long history of states’ reform efforts has generated substantial variation in the
explanatory variable both across time and across states, facilitating powerful (if
complicated) models.
Third, organizations have provided funding to support malpractice research.
The Robert Wood Johnson Foundation (RWJF), for example, has made a major
financial commitment in this area.195 The fact that the topic has clear implications for core areas of health services research, such as the financing and quality
of health care, has made it possible for researchers to compete for funding under
general solicitations for health policy and health services research projects as
well. The availability of extramural funding has made it possible for scholars in
soft-money environments to enter the field and for researchers to carry out large
data-collection efforts.
Less evident are the enabling factors that allow for the use of this research in
malpractice reform efforts. Although academics have sometimes been invited to
give testimony in state and federal hearings and to brief legislative staff, they
have not had the kind of access that lobbyists for stakeholder organizations have
Medical Liability System, 82 N.C. L. REV. 759 (2004)). Yet, the court utterly dismissed a controlled,
multivariate analysis by a well-regarded health economist on the basis that it included states that capped
total damages as well as states that capped only noneconomic damages. See id. at 505 (citing Kenneth
E. Thorpe, The Medical Malpractice ‘Crisis’: Recent Trends and the Impact of State Tort Reforms, 23
HEALTH AFF. w4–20 (2004), available at http://content.healthaffairs.org/cgi/reprint/hlthaff.w4.20v1.pdf.
The inclusion of those four states may have affected the study findings, but the Thorpe study certainly
still had more explanatory power than the other cited sources. Moreover, the court omitted any mention
of a large number of other controlled studies. See MELLO, supra note 165, at 24–25 (collecting studies).
195. See ROBERT WOOD JOHNSON FOUND., IMPROVING MALPRACTICE PREVENTION AND COMPENSATION
SYSTEMS (2002), available at http://www.rwjf.org/reports/npreports/impacs.htm. More recently, the
Foundation has funded malpractice research out of its Pioneer Portfolio and its Health Care Financing
and Organization program.
688
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
had.
One way to counter this problem is to make EHL research more accessible to
the policy community. Perhaps the most important means to this end has been
the dedication of research sponsors and organizations to generating and disseminating high-quality malpractice research for consumption by policymakers.
RWJF and AcademyHealth have been at the leading edge of these efforts,
employing skilled communications staff to translate complex research findings
into easily digestible policy briefs that are then aggressively disseminated to the
policy community.196 RWJF’s Synthesis Project, for instance, “produce[s] concise and thought-provoking briefs and reports that translate research findings on
perennial health policy questions” by “synthesizing what is known, while
weighing the strength of the research evidence and exposing gaps in current
knowledge.”197 Indeed, the Synthesis Project engaged one of the authors of this
Article to synthesize the literature on tort reforms, and based on this analysis,
Project members created briefs and forums for the dissemination of the results,
and made the products freely downloadable on the internet.198 Few other areas
of empirical health law research have received so great a commitment to
dissemination.
A number of challenges have made it difficult for EHL research to achieve its
scientific and policy aims, however. Many of the richest datasets on claims exist
at the insurer or state level and are not clearly generalizable or integrable with
one another. All existing datasets that have supported secondary analyses have
significant limitations. For example, jury verdicts exclude settlements; the Data
Bank excludes unpaid claims; and state closed-claim databases omit unpaid
claims and some kinds of health care providers, may have suboptimal data
accuracy, and include reporting fields that may vary over time.199 The best data
remain proprietary: the records of liability insurers and the providers they
insure. These data are difficult to gain access to and cumbersome and expensive
to extract. This, too, leads to datasets that are limited in geographic scope; it is
not practicable to review thousands of litigation records in every state. When
only a few states are represented, there are very limited possibilities for
evaluating state tort reforms. For this reason, although studies based on insurers’ litigation records and/or medical chart review have made seminal contributions to describing the epidemiology of medical injury and malpractice
196. See, e.g., Christal Stone, Findings Brief: Investment Returns and Size of Damage Caps Impact
Rising Cost of Malpractice Premiums, ACADEMYHEALTH, Apr. 2007, at 1, available at http://www.hcfo.net/
pdf/findings0407.pdf.
197. Robert Wood Johnson Found., About the Project, SYNTHESIS PROJECT (2004), http://www.rwjf.org/
pr/synthesis/about_the_project/index.html (last visited Oct. 22, 2007).
198. These works are collected at Robert Wood Johnson Found., Medical Malpractice: Impact of the
Crisis and Effect of State Tort Reforms, SYNTHESIS PROJECT (2006), http://www.rwjf.org/pr/synthesis/
reports_and_briefs/issue10.html (last visited Oct. 22, 2007).
199. AM. ACAD. OF ACTUARIES, IMPORTANT CONSIDERATIONS WHEN ANALYZING MEDICAL MALPRACTICE
INSURANCE CLOSED-CLAIM DATABASES (2005), available at http://www.actuary.org/pdf/casualty/
medmal_042005.pdf.
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
689
claiming,200 they have not evaluated tort reforms.
There are also real methodological challenges to modeling the complex
tort-reform environment. Controlling for other state characteristics which may
affect the outcome variables can be difficult, and fixed-effects models are not
feasible for some periods when tort reforms were time invariant in nearly all
states. There are also concerns about endogeneity, because tort reforms are most
likely to be passed by states that are experiencing volatility in insurance
premiums or claims costs, and there are questions about how to construct
variables to represent a large and heterogeneous group of reforms.
The main barrier to the productive use of malpractice research in policymaking has been that the involved political interest groups have not played a
facilitative role. Throughout the past three malpractice crises, the policy debate
has been dominated by two interest groups—trial lawyers and health care
providers—and their allies.201 Their focus has been single-minded: Are noneconomic damages caps a good idea or a bad idea?202 Although empirical legal
scholars have repeatedly tried to shift policymakers’ focus to reforms that are
more likely to address fundamental and well-documented problems with the
medical liability system, few interest groups have followed their lead.203 Groups
have instead homed in on the narrow, if important, problem of spiraling liability
costs and the reform that most directly addresses the size of malpractice
awards.204
These groups have been fiercely adversarial in their postures and have
appeared to believe that maintaining those postures requires the utter rejection
of empirical evidence that does not support their position. This has resulted in a
highly polarized debate with none of the nuance that characterizes evidencebased policy making,205 and in which “[e]ach side portrays the other in very
harsh language and portrays the other side’s position as disastrous for society.”206 Rather than nurturing policymakers’ interest in data, these interest
groups have tended to strongly emphasize anecdotes featuring sympathetic
victims of malpractice and the malpractice crisis. These narratives have strong
appeal to legislators and journalists because they are easy to understand and
200. The best-known such study is the Harvard Medical Practice Study, discussed supra note 60.
201. See Kersh, supra note 151, at 54–55, for a historical account of this adversarial policy debate.
202. William M. Sage, Malpractice Reform as a Health Policy Problem, in MEDICAL MALPRACTICE IN
THE U.S., supra note 151, at 30, 31 (“Both long-standing antipathy between doctors and lawyers and the
macro-politics of tort law tend to fixate the malpractice debate on lowest-common-denominator reforms
of the legal process that are familiar to the medical profession and that translate readily to nonmedical
litigation.”).
203. William M. Sage, Medical Malpractice Insurance and the Emperor’s Clothes, 54 DEPAUL L.
REV. 463, 463–64 (2005).
204. Robert L. Rabin, The Politics of Tort Reform, 26 VAL. U. L. REV. 709, 715–16 (1992).
205. Neil Vidmar et al., Uncovering the “Invisible” Profile of Medical Malpractice Litigation:
Insights from Florida, 54 DEPAUL L. REV. 315, 317 (2005) (“Angry rhetoric and poorly substantiated
claims often substitute for systematic data . . . .”).
206. Anthony J. Sebok, Dispatches from the Tort Wars, 85 TEX. L. REV. 1465, 1466 (2007) (book
review).
690
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
have a human face.207 Statistical studies, in contrast, require a great deal of
cognitive work and often lead to a grey, rather than black or white, view of the
world.
Scholars have not had the incentives that interest groups have had to make
participation in and influence over the policy debate a priority.208 While professors have struggled to find time and money to attend briefings and hearings,
well-financed advocates have been working full-time to press their agendas in
those settings and through extensive behind-the-scenes lobbying.209 It is not
surprising that it is their position papers that get read, nor that these rarely
include data from scholarly studies.
A second problem has been that over time the evidentiary field has become
quite crowded. The number of analyses by academics, government agencies,
and political interest groups and their consultants has skyrocketed during the
latest malpractice crises. There are now so many studies of tort reforms that it
requires considerable effort to find and review them all, much less evaluate
which are of good or poor quality. In other areas of health policy, communities
of scientists have come together to rigorously catalog and analyze the literature
on topics of great importance for health care providers and policymakers, but
there are few such efforts in health law.210
The problem of evaluation is exacerbated by the great complexity of many
academic studies of tort reforms. Some legislators struggle with even basic
statistical concepts,211 and as methodological innovations in econometrics have
led to ever more sophisticated analyses, it has become increasingly difficult for
207. Id. at 1500.
208. HALTOM & MCCANN, supra note 182, at 101. Haltom and McCann also note that scholars’ “very
claim to authority as professional experts has limited the scholars’ public visibility,” id. at 17,
presumably because it requires a heavy investment of time to maintain the necessary base of expertise,
and scholars may consider it unseemly to descend into political fights rather than focus on perfecting
their craft.
209. Sebok, supra note 206, at 1501 (“The sociolegal scholars, as full-time academics, had neither
the time nor the resources to promote themselves like the tort reformers at the Manhattan Institute.”
(citing HALTOM & MCCANN, supra note 182, at 101)).
210. The Cochrane and Campbell Collaborations conduct systematic reviews of the evidence on
matters of clinical care and select areas of social policy (including a few public health interventions),
respectively. See Campbell Collaboration, About the Campbell Collaboration, http://www.campbell
collaboration.org/About.asp (last visited Oct. 22, 2007); Cochrane Collaboration, Cochrane Reviews
and the Cochrane Library—An Introduction, http://www.cochrane.org/reviews/clibintro.htm (last visited Oct. 22, 2007). Although systematic reviews potentially are useful for policymakers in deciding
which health care services should be funded by government programs, this method is not commonly
used to study the impacts of specific legal rules on outcomes.
211. An example is a member of Congress’s attempt to cross-examine a witness at a committee
hearing using data from the National Practitioner Data Bank. The witness, an insurance association
representative, had offered insurer data indicating that the mean malpractice payout was $496,000. Rep.
Bart Stupak asserted that that figure did not “jibe” with Data Bank data indicating that the median
payout was $125,000 and appeared not to understand the witness’s explanation that the mean and
median are two different statistics. See Assessing the Need To Enact Medical Liability Reform: Hearing
Before the Subcomm. on Health of the H. Comm. on Energy and Commerce, 108th Cong. 145–46
(2003) (testimony of Lawrence E. Smarr, President, Physicians Insurers Association of America).
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
691
non-experts to decipher and critically evaluate their findings. Adversaries in the
policy debate use this to their advantage by attempting to discredit the studies,212 leading to further confusion.
Finally, the way in which malpractice has been framed as a policy issue has
inhibited deliberative lawmaking. Framing the problem of rising liability insurance costs as a “crisis,”213 however defensible from the perspective of those
who have to pay those costs, has driven policymakers toward a narrow goal—
stabilizing premiums—and a narrow solution—damages caps. The atmosphere
of emergency has made it difficult to engage policymakers in deliberations
about persistent, fundamental problems with the malpractice system and farreaching reforms that could address those problems along with the problem of
insurance costs.214 Rather, policymakers have been attracted to seemingly
simple fixes that can satisfy powerful interest groups’ demands to do something.215 It is one of the great paradoxes of malpractice politics that the
conditions that maximize the chances that legislators can mobilize enough
support to pass legislation also minimize the chances that the legislation will
reflect careful, thorough consideration of what the best evidence suggests will
be a successful reform. Together, these factors explain why empirical research
on malpractice reforms has not lived up to its promise for shaping policy.
C. SUMMARY
Viewed together, the two case studies help to identify a common set of
factors that affect the likelihood EHL research will influence policy decisions.
In general, EHL research is more plentiful and impactful when (1) there is a
considerable amount of variation in the health laws in question across jurisdictions and over time, which strengthens researchers’ ability to statistically estimate the effects of the laws; (2) data on laws, their outcomes, and other factors
that might confound analysis of the relationship between the laws and the
outcomes are readily available; (3) the issue is convincingly framed as a public
health issue and visible enough to attract skilled empirical researchers and
212. See, e.g., Don Finley, Study Says Medical Malpractice Crisis Is a Myth, SAN ANTONIO
EXPRESS-NEWS, Mar. 11, 2005, at 3B (quoting Dr. Bohn D. Allen, President of the Texas Medical
Association, as saying, “The authors of this study used data that was inflation-adjusted, populationadjusted, and health-care-cost adjusted. It was so adjusted that it became truth-adjusted.”). The study to
which Allen referred reported time trends in medical malpractice filing rates and payouts size. The
authors wished to examine whether out-of-control juries were behind increases in medical liability
insurance premiums; therefore, they adjusted their estimates to eliminate obvious alternative explanations, including increases in damages due to rising health care costs and general inflation and increases
in the quantity of medical services provided due to population growth. See Bernard Black et al.,
Stability, Not Crisis: Medical Malpractice Claim Outcomes in Texas, 1988–2002, 2 J. EMPIRICAL LEGAL
STUD. 207 (2005).
213. See Kersh, supra note 151, at 50 (discussing the influence of framing on the politics of
malpractice reform).
214. See TOM BAKER, THE MEDICAL MALPRACTICE MYTH 180 (2005).
215. Michelle M. Mello, Managing Malpractice Crises: Why Is It So Hard?, 33 J.L. MED. & ETHICS
414 (2005).
692
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
funders of both primary studies and systematic reviews; (4) infrastructures exist
within the academy to support the production and dissemination of empirical
research; (5) interest groups act as facilitators rather than barriers to basing
policy decisions on the best available evidence; and (6) policymakers understand and have confidence in the methods used to produce relevant empirical
findings.
When these stars do not align, empirical research, if any exists, has less
impact on policy decisions. Although some of the factors surely are out of the
control of the legal academy, steps can be taken inside the academy to increase
the likelihood that the enabling conditions are satisfied. In the following Part we
offer specific suggestions for ways in which the legal academy can both
increase the production of high-quality EHL studies and increase the impact of
these studies on the implementation of health law and policy.
IV. FUTURE DIRECTIONS
The case studies examined in Part III serve as a useful jumping-off point to
consider more generally the barriers that impede the production of EHL research and limit its impact on health law and policy. While many impediments
are common to EHL research conducted both within and outside the legal
academy, several are unique to legal academia. Pushing the field forward will
require that these barriers be addressed. We offer a few suggestions along these
lines, some of which emanate from the case studies we have analyzed.
A. WAYS TO INCREASE PRODUCTION OF HIGH-QUALITY EHL STUDIES
Although production of EHL studies has increased substantially over the past
three decades, as evidenced in the two cases discussed in Part III, a broader
view of the field suggests that much more research is needed to develop a fuller
picture of how health law and policy influence outcomes. Several reasons
account for the gaps in knowledge and the fact that a relatively small portion of
this literature is produced by scholars within the legal academy. In particular,
production of EHL studies is hindered by a lack of easy and inexpensive access
to reliable data, institutional barriers, and a general lack of acceptance of the
scientific approach within law schools. These inhibiting factors present challenges, but interested parties within the legal academy can take specific steps to
break down these barriers and increase the number of high-quality EHL studies
produced by legal scholars. In particular, they can work to attract more funding
to obtain data, encourage collaborations between legal scholars and methodologists, and educate law students and faculty on empirical methods and the merits
of employing a scientific approach.
1. Increase Access to and Collection of Reliable Data
As we discussed in Part II.B., a major factor inhibiting the production of EHL
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
693
research is the difficulty of accessing reliable data.216 Although EHL studies
often employ data that directly measure the variables of interest, empirical
findings are often based on proxies—variables the researchers hope are positively correlated with the true variable of interest—because data on the variables of interest are not available or are too expensive to collect. In the most
troubling cases, connections between legal rules and outcomes go unexplored
because data allowing either direct measurements or proxy-based measurements
are unavailable.
For EHL research to evolve, availability of data and data-collection methods
need to evolve. We offer two suggestions. First, EHL scholars should more
aggressively seek out funding to collect or obtain data not currently available.217 Research sponsors will be more likely to view EHL study proposals
favorably if these proposals are not anomalies among the submissions they
receive, and if the proposals make a strong case that such research is important
and can make a difference to public health. As we increase our production of
impactful research, it will become easier to demonstrate this.
Second, scholars should also encourage government agencies that collect data
for regulatory purposes, as well as private organizations in the health care
market, to make their data more available and useful to researchers.218 These
entities should move toward collecting data with the same contemplation of
their possible research uses as occurs in other fields, such as labor and health
care. We need to move from a world where researchers peck around at datasets
collected for other purposes to a world where data are collected systematically
with research in mind. We also need to better address agencies’ and companies’
concerns about data confidentiality and security. Methods widely in use in other
areas of health research, such as development of limited, deidentified (or
partially deidentified) datasets for public use and execution of data-use agreements, could be applied here.
2. Break Down Institutional Barriers
Legal scholarship traditionally has been produced under a model emphasizing
theoretical analysis and solo authorship.219 Although it has begun to expand
outside this model, general adherence to and preference for the traditional
216. See, e.g., Deborah L. Rhode, Legal Scholarship, 115 HARV. L. REV. 1327, 1353–54 (2002)
(discussing difficulties with data collection and other obstacles to producing empirical legal studies).
217. See id. at 1360 (“Many academics are strikingly clueless about how to market their scholarship
to funders and the media. Writing effective grant proposals and press releases is an acquired skill—and
one not taught in law schools. It should be taught, at least to faculty.”).
218. One recent example is the involvement of academics in the debate over the National Association of Insurance Commissioners’ (NAIC) model law requiring disclosure by medical malpractice
insurers of closed-claim data. See NAIC Statistical Info. Task Force, http://www.naic.org/committees_
c_sitf.htm (last visited Oct. 22, 2007).
219. See Tracey E. George & Chris Guthrie, Joining Forces: The Role of Collaboration in the
Development of Legal Thought, 52 J. LEGAL EDUC. 559, 560 (2002) (“[M]ost law review articles seem
to be written by one author.”); Heise, supra note 9, at 812 (“[T]he amount of theoretical and doctrinal
694
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
approach has created institutional barriers within the legal academy to the
production of empirical legal research.220 Because health law enjoys little
prestige among legal scholars, EHL studies are especially marginalized, as these
studies lie at the intersection of two nontraditional areas of scholarship.221
One challenge is overcoming the traditional model’s low esteem for collaborative research.222 A large portion of the most compelling EHL research has been
produced though collaborations between legal scholars and scholars from other
fields who contribute disciplinary grounding and methodological expertise.
Collaborations are more likely to form when academic environments create
incentives for faculty to seek them out (for example, in “soft money” settings,
collaborations are essential because strong teams have a competitive advantage
in grant seeking). At present, these incentives do not exist in law schools, and
the emphasis on sole-authored work in the promotion process is a strong
disincentive to engage in collaborative work.223 There are signs of change in
some law schools—for example, establishing centers for health law research
with faculty affiliates from around the university,224 sponsoring interdisciplinary
workshops and conferences,225 and providing grants for collaborative research.226 We would submit, however, that the balance of incentives will not tip
until coauthored work comes to enjoy greater legitimacy.227
Preference for the traditional model has afforded less prestige to the use of
empirical, rather than theoretical, research approaches.228 Various factors acscholarship that presently fills many law review pages overwhelms the amount of empirical scholarship.”).
220. Thomas S. Ulen, The Unexpected Guest: Law and Economics, Law and Other Cognate
Disciplines, and the Future of Legal Scholarship, 79 CHI.-KENT L. REV. 403, 413 (2004) (“[T]here
seems to be a widely held belief in the legal academy that the scientific method is inappropriate for the
study of law.”).
221. See Mark A. Hall & Carl E. Schneider, Where Is the “There” in Health Law? Can It Become a
Coherent Field?, 14 HEALTH MATRIX 101 (2004).
222. George & Guthrie, supra note 219, at 571 (“Collectively, the comprehensive citation-count
studies demonstrate that articles by coauthors have had less impact than we would have expected given
the rate at which coauthored articles have been published.”).
223. See id. at 560 (“[T]here are practical impediments to collaboration in law schools that may not
exist in law practice. For example, law schools often give faculty only token credit for coauthored
works.”).
224. Harvard Law School’s Petrie-Flom Center for Health Law Policy, Biotechnology and Bioethics
and Georgetown University Law Center’s O’Neill Institute for National and Global Health Law are two
recently established examples.
225. The Petrie-Flom Center has sponsored a number of such conferences and runs an annual
semester-long health law and policy workshop.
226. For example, Georgetown University Law Center provides research grants for collaborative
research to cover costs of travel and other expenses.
227. Others have suggested collaborations as one way to help mitigate the difficulties in meshing
legal scholarship and social science. See, e.g., Jeremy A. Blumenthal, Law and Social Science in the
Twenty-First Century, 12 S. CAL. INTERDISC. L.J. 1, 37–38 (2002).
228. See Robert W. Gordon, Lawyers, Scholars, and the “Middle Ground,” 91 MICH. L. REV. 2075,
2085 (1993) (“[Empirical legal research] remains to this day the most neglected and ridiculously
undervalued as well as the most potentially fruitful branch of legal studies.”); Craig Allen Nard,
Empirical Legal Scholarship: Reestablishing a Dialogue Between the Academy and Profession, 30
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
695
count for this disparity, but chief among them are a lack of familiarity with
empirical methods, a lack of appreciation for the ways in which observation and
measurement can enrich (rather than replace) theory, and an understanding of
the concept of intellectual creativity that focuses on the development of innovative ideas rather than discoveries about the world. Closing this gap requires
educating legal scholars in empirical methods, clarifying the role of empirical
research in theory development, demonstrating the impact of empirical research
on policy and human welfare, and nurturing the idea that such real-world impact
is an important measure of the excellence of scholarship. As it becomes clearer
that the usefulness of theory hinges, in part, on whether a theory’s predictions
are consistent with observations from the field or the laboratory, legal scholars
will come to appreciate more fully the value of empirical work. As more law
schools offer courses in empirical research methods,229 the cadre of scholars
able to conduct empirical research will grow.
As familiarity with and appreciation for empirical work grows, the preferred
model of legal scholarship will continue to evolve. Furthermore, as empirical
and collaborative research becomes more common, legal scholars will gain
experience evaluating the work for purposes of hiring and promotion. This will
further break down the barriers that discourage legal scholars from engaging in
this type of research.230
Of course, the proposals set out here are not meant to imply that empirical
studies should replace non-empirical research or that resources should be
shifted away from non-empirical research. Clearly, empirical methods are not
well suited for all important investigations. Empirical research depends in many
cases on theoretical and doctrinal research for grounding.231 Although a large
portion of EHL research is purely descriptive, much of it is performed to test
theoretical predictions and assumptions about causal connections on which
theorists and policymakers commonly rely. Our intent is to argue for an increase
in the production of high-quality empirical research while at the same time
preserving the important role of and respect for theoretical and doctrinal
analysis.
WAKE FOREST L. REV. 347, 361–65 (1995) (offering reasons why empirical legal research has taken a
back seat to other forms of legal scholarship).
229. See Theodore Eisenberg, Why Do Empirical Legal Scholarship?, 41 SAN DIEGO L. REV. 1741,
1742 & n.3 (2004) (mentioning a short course for law professors offered at Washington University-St.
Louis School of Law and a number of other such programs at various law schools); Howell E. Jackson,
Analytical Methods for Lawyers, 53 J. LEGAL EDUC. 321 (2003) (discussing methods courses for law
students).
230. See Nard, supra note 228, at 363 (finding that “twenty-five percent of non-tenured professors
were concerned that empirical research would not be viewed favorably by their respective tenure
committees”); Ulen, supra note 220, at 418 (“Junior law faculty may perceive making a commitment to
empirical techniques as a high-risk, and, therefore, unattractive strategy for promotion . . . .”).
231. See generally Thomas S. Ulen, A Nobel Prize in Legal Science: Theory, Empirical Work, and
the Scientific Method in the Study of Law, 2002 U. ILL. L. REV. 875 (describing the relationship between
theory and empirical work and arguing the benefits of using the scientific method to study law).
696
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
B. WAYS TO INCREASE THE IMPACT OF EHL STUDIES ON HEALTH LAW AND POLICY
Increasing production of EHL studies is but a first step. These studies create
minimal value if they have little or no impact on the structure of law and the
shape of policy. As the case studies presented earlier demonstrate, high-quality
EHL studies do not always impact law and policy to the extent they should.
Several reasons account for this, including lack of rigor, failure to aggressively
disseminate new findings, and difficulties policymakers have in reading, interpreting, and critiquing empirical studies. As with the production barriers, interested
parties within the academy can take steps to mitigate or eliminate these barriers.
1. Strengthen Rigor
Although much EHL research can be classified as rigorous according to
scientific standards, this is not uniformly the case. As EHL scholarship produced by legal scholars continues to converge toward social-science research
with respect to the methodology employed and the presentation of results,
publishers and consumers of EHL research will demand that it meet the
standards of scientific rigor to which other kinds of social-science research are
held.232 The legal academy can help strengthen the rigor of EHL studies by
fine-tuning hiring and tenure standards and making greater use of peer-review
processes.
Law schools traditionally have not been in the business of hiring empiricists;
therefore, law faculties have relatively little experience evaluating the quality of
empirical studies produced by job candidates. Promotion committees and processes also do not make much effort to understand the dynamics of collaborative research, and thus may misapprehend or underappreciate the unique
intellectual contributions that faculty have made in collaborative projects. Even
such basic matters as how the order of authors listed on articles in different
disciplines reflects the authors’ contributions may be mysterious. Constituting
hiring and promotion committees such that candidates know that good-quality
research and important contributions will be recognized and rewarded, and that
sloppy studies and free-riding in collaborations will be duly noted in the
committees’ decisions would do a great deal to spur the kind of EHL research
we want and need. Involving reviewers from other departments within the
university would be a low-cost way to bring in additional expertise. Hiring
committees can also encourage the growth of EHL research by favoring candidates who have formal methods training or a demonstrated record of engaging
in collaborations with methodologists that produce high-quality, impactful EHL
232. Although this section contains suggestions for interventions to increase the rigor of empirical
health law research, some have argued that the large and growing number of empiricists and ready
availability of data will allow the public easily to debunk “outlandish or mistaken claims.” See, e.g.,
McGinnis, supra note 8, at 51 (“[T]he large number of empiricists will create a rich market for
empirical work and provide greater incentives to get the facts and models absolutely correct. There are
now so many investigators and the computer technology is so pervasive that results are constantly
checked.”).
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
697
studies.
Consideration should also be given as to how to more fully reap the advantages of peer-review processes to improve empirical legal scholarship. The
formal standards (for example, criteria for authorship) and informal norms (for
example, minimum response rates for surveys) of peer-reviewed journals in
other fields may be appropriate to apply to some kinds of EHL studies. The
most rigorous journals require authors to submit data and logs of procedures
used to derive reported results.233 After the paper is published, some journals
post the data and result logs on their websites.234 These procedures make
replication and further investigation easier for other researchers.
EHL scholars should also be encouraged to submit their work to peerreviewed journals in both the legal and health fields and serve as referees for
these journals.235 Undergoing the peer-review process, though time consuming
and often burdensome for authors, can markedly improve the quality of a
research paper, as well as facilitating the match between the quality of the
research and the prestige of the journal in which it appears. This match is
important because journal reputation has a signaling effect for consumers of
research, indicating the quality and reliability of the research published there.
We readily concede that peer-review processes and their results are far from
perfect,236 but submit that they are more reliable than the methods by which
most student-run law journals screen articles for publication.237
Law reviews and journals can also take steps to ensure the rigor and
reliability of results they publish. They can request reviews by faculty with
empirical methods training for portions of submitted articles that student editors
are not qualified to evaluate. The journals also can require submission of data
and results logs or encourage authors to post this information on their own
websites.238
233. See, e.g., American Economic Review: Data Availability Policy, http://www.aeaweb.org/aer/
data_availability_policy.html (last visited Oct. 22, 2007) (requiring authors to submit “data, programs,
and other details of the computations sufficient to permit replication”).
234. See, e.g., id. (noting the posting of the authors’ submitted information).
235. See Gregory Mitchell, Empirical Legal Scholarship as Scientific Dialogue, 83 N.C. L. REV. 167
(2004) (suggesting a disclosure system that law reviews can implement to increase critical peer review
of empirical legal research). Our malpractice case study might suggest, however, that this might limit
access of the research to judges who seem to use primarily legal databases and internet searches. Public
access to peer-reviewed journals, though, is improving as more make their full-text articles available
online.
236. See Frank B. Cross, The Naı̈ve Environmentalist, 53 CASE W. RES. L. REV. 477, 486 (2002)
(“Peer review is demonstrably unreliable at screening research for validity. It tends to be infected by
ideological biases and replicate the preferences of the editor and reviewers. Simply because something
is peer-reviewed does not make it true, nor does the absence of peer review make information false.”).
237. See Rhode, supra note 216, at 1356 (discussing problems with student-run review processes);
Schneider & Teitelbaum, supra note 8, at 103–04 (same).
238. Cf. Mitchell, supra note 235, at 176 (recommending that law reviews adopt a “set of stringent
disclosure requirements” rather than a peer-review process). Authors can also voluntarily post datasets
and result logs on their own websites for public access. For an example, see John Donohue’s website
698
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
2. Increasing Widespread and Effective Dissemination of New Findings
Increasing the quality of EHL studies is the first step toward increasing their
impact on policy. Once produced, however, the studies will have little influence
on policy unless the results end up in the hands of policymakers. The first
obvious step to disseminate findings is publication. As the New York Times
recently reported, however, legislators and judges might not read scholarly
journals as often as we might hope.239 Therefore, we need to take additional
steps to increase our impact on policy debates.
Medical and public health schools have established dedicated personnel in
offices of communications to disseminate faculty research findings. For example, faculty members at the Harvard School of Public Health have a press
officer assigned to them who becomes familiar with their portfolio of work.240
When a new paper is about to be published, faculty can send the galleys to their
officer and discuss the message the authors would like to convey. If appropriate,
the press officer then drafts a press release, contacts reporters who might be
interested in the work, and arranges interviews. He also disseminates the
research on the school’s website and through podcasts and listservs. These
arrangements are expensive, but highly effective, because they utilize media
professionals and minimize demands on faculty members’ time. Building such
structures in law schools would do a great deal to raise the public profile of
EHL research.
Widespread dissemination is necessary for EHL researchers to impact policy
but may not be sufficient. A second measure that our case studies clearly
suggest would be helpful is for scholars to engage in efforts to make their
empirical research more comprehensible to policymakers and judges.241 Legal
scholars now have more to offer to policymakers than mere learned opinion, but
scientific results can be difficult to explain and to understand and can be
challenged on many different bases. By describing results in a manner that law
audiences can understand and providing clear warnings regarding limitations,
researchers can help policymakers import the studies into debates in appropri-
providing data and result logs for several empirical publications at http://islandia.law.yale.edu/donohue/
pubsdata.htm (last visited Oct. 22, 2007).
239. Adam Liptak, When Rendering Decisions, Judges Are Finding Law Reviews Irrelevant, N.Y.
TIMES, Mar. 19, 2007, at A8. Rhode has remarked that “[a]cademics write mostly for one another and
for others in their profession. Seldom do we [consciously] seek a mass audience, and rarely do we find
one.” Deborah L. Rhode, A Bad Press on Bad Lawyers: The Media Sees Research, Research Sees the
Media, in SOCIAL SCIENCE, SOCIAL POLICY, AND THE LAW 139, 140 (Patricia Ewick et al. eds., 1999),
quoted in HALTOM & MCCANN, supra note 182, at 73; see also Michael D. McClintock, The Declining
Use of Legal Scholarship by Courts: An Empirical Study, 51 OKLA. L. REV. 659 (1998).
240. See Harvard School of Public Health, Office of Communications, http://www.hsph.harvard.edu/
administrative-offices/communications/ (last visited Oct. 22, 2007).
241. For a useful discussion of the history of attempts to inform policymakers about empirical
findings, see Daniel M. Fox & Lee Greenfield, Helping Public Officials Use Research Evaluating
Healthcare, 14 J.L. & POL’Y 531 (2006).
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
699
ate, useful, and defendable ways.242 They might also consider creating “policymakers’ digests”—short issue briefs accompanying research findings relevant to
active policy debates. In health research, study sponsors routinely produce such
briefs243 and disseminate them to policymakers, making liberal use of electronic
newsletters as well. Finally, participation in legislative briefings and hearings,
though time-consuming and subject to receipt of an invitation, can be an
important means of clarifying the message of one’s research; the motor vehicle
safety law case study suggests that such participation can be effective in
shaping policy. Similarly, scholars can help ensure that study findings are not
“spun” too much for political purposes by responding to such misuse through
op-eds and other outlets.
The story of medical malpractice reform also indicates that for some policy
issues, it is critical for scholars to help policymakers sort through a large body
of conflicting empirical analyses of varied quality, fit individual studies into the
larger policy picture, and understand what policy prescriptions emerge from the
empirical findings. Otherwise, there is a danger that they will not rely on the
best available evidence or will draw faulty implications from the results of
narrow studies.244 One way to assist policymakers is to publish literature
reviews with an emphasis on determining what conclusions can be drawn from
the available body of evidence and identifying areas where there is insufficient
data to draw reliable conclusions.245
In addition, systematic reviews modeled on structured methods would be
invaluable in helping policymakers draw conclusions from bodies of EHL
research. Such reviews have been powerful and useful tools for advancing
clinical and public health practice. The reviews produced by the Cochrane
Collaboration and Campbell Collaboration, for example, rigorously evaluate the
available evidence regarding a clinical issue or social program, respectively. 246
Their rigor is such that they are widely recognized as the definitive statements
of the evidence on a particular issue. Similarly, the Centers for Disease Control
and Prevention has convened an independent Task Force on Community Preventative Services to produce systematic reviews of primary research studying the
242. Id. at 538; see also Lee Epstein et al., On the Effective Communication of the Results of
Empirical Studies (pt. 1), 59 VAND. L. REV. 1811 (2006). To reach judges, further measures may be
required, such as greater representation of empirical scholars in amicus briefs.
243. See, e.g., The Synthesis Project, http://www.rwjf.org/pr/synthesis/index.html (last visited Oct.
22, 2007) (collection of recent Synthesis Project briefs).
244. See Christal Stone, Issue Brief: Medical Malpractice: Strengthening the Evidence Base,
ACADEMYHEALTH, Dec. 2006, at 1, 1, available at http://www.hcfo.net/pdf/issue1207.pdf (last visited
Oct. 22, 2007).
245. See, e.g., MELLO, supra note 165; see also Randall R. Bovbjerg & Robert A. Berenson,
Surmounting Myths and Mindsets in Medical Malpractice, URB. INST. HEALTH POL’Y BRIEFS, Oct. 2005,
at 1, available at http://www.urban.org/UploadedPDF/411227_medical_malpractice.pdf (identifying
sound conclusions and “myths” in the medical malpractice policy debate). Because many courts look
only to law reviews, EHL scholars should also aim to publish literature summaries in law reviews. As
courts gain the capacity to search additional databases, publication venue will become less of a concern.
246. Campbell Collaboration, supra note 210; Cochrane Collaboration, supra note 210.
700
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
effectiveness, economic efficiency, and feasibility of interventions to promote
community health and disease prevention.247 Among the interventions evaluated
to date have been public health laws and policies aimed at reducing tobacco
use,248 encouraging physical activity,249 and reducing firearm injuries.250 Such
evaluations could be much more widely applied to health laws.
A second way researchers can help is to explicitly bridge individual study
results into the policy debates. Primary EHL studies usually are insufficiently
broad to generate specific policy recommendations because they tend to focus
on narrow components of multi-faceted issues. By including a discussion of the
implications individual results have for the larger policy debate, researchers can
help policymakers understand the power and the limitations of narrow results.
How can law schools facilitate these initiatives? Again, tweaking the incentive structure for faculty is key. Hiring and tenure standards can be adjusted to
encourage empirical scholars to take active steps to inject their research into
policy debates. Current standards emphasize a candidate’s publication record
and number of citations to the candidate’s work by other legal scholars over
evidence that a candidate’s scholarship has been read, cited, and relied upon by
policymakers and judges. Readiness for promotion should be evaluated not only
by reference to how the individual’s work has advanced her field of scholarship,
but also by how it has impacted the world outside the legal academy. Adding
this standard to the criteria for promotion and tenure decisions will create
incentives for empirical scholars both to increase the rigor of their work and to
disseminate their findings widely.
3. Educating Consumers of EHL Research
Finally, in addition to strengthening the rigor of EHL research and improving
dissemination and communication, we can increase the impact of our research
by helping to educate potential audiences for EHL about how to read, interpret,
and evaluate empirical studies. Some in the legal academy have been engaged
in this endeavor for many years,251 but as the production of empirical studies
247. See Task Force on Cmty. Preventative Servs., The Community Guide: Overview, http://
www.thecommunityguide.org/about/default.htm (last visited Oct. 22, 2007).
248. See Task Force on Cmty. Preventative Servs., Effectiveness of Smoking Bans and Restrictions
To Reduce Exposure to Environmental Tobacco Smoke (ETS) (Jan. 3, 2003), available at http://
www.thecommunityguide.org/tobacco/tobac-int-smoke-bans.pdf.
249. See Task Force on Cmty. Preventative Servs., Creating or Improving Access to Places for
Physical Activity Is Recommended To Increase Physical Activity (Nov. 15, 2005), available at http://
thecommunityguide.org/pa/pa-int-create-access.pdf.
250. See Task Force on Cmty. Preventative Servs., Review Finds Insufficient Evidence To Assess the
Effects of Eight Firearms Laws on Violence (Sept. 9, 2003), available at http://www.thecommunityguide.
org/violence/viol-int-firearms.pdf.
251. See, e.g., Fox & Greenfield, supra note 241, at 538–50 (describing methods used to educate
both public officials and members of the judiciary on health services effectiveness research methods);
Daniel L. Rubinfeld, Guide to Multiple Regression, in 1 MODERN SCIENTIFIC EVIDENCE: THE LAW AND
SCIENCE OF EXPERT TESTIMONY 147, 147–83 (David L. Faigman et al. eds., 1997) (educating lawyers
about regression analysis); Daniel L. Rubinfeld, Reference Guide on Multiple Regression, in REFERENCE
2008]
EMPIRICAL HEALTH LAW SCHOLARSHIP
701
increases, we likely will see an increase in the demand for education on the part
of legislators, legislative staff, judges and clerks, and law faculty.
Because lawyers dominate all of these groups of constituents for EHL
research, one approach is to include this type of training as part of the standard
legal education curriculum.252 While many law schools offer quantitativemethods courses, they are not universally available, much less required. Those
that do not offer such courses should consider hiring faculty or adjunct staff to
teach them. Short courses, similar to those offered to train judges in scientific
research methods so that they can serve as quality gatekeepers at trial, can also
be an effective way to give jurists a basic familiarity with empirical research
and how to find it.
The politics of health policy decisions will continue to create pressures for
lawmakers to focus on the desires of special interest groups and their lobbyists
even when sound empirical evidence counsels that policy should move in a
different direction. For some issues and some policymakers, increases in the
funding, volume, visibility, quality, comprehensibility, and accessibility of empirical research will be insufficient to change this focus. But addressing these
conditions will reduce the time costs and cognitive barriers, if not the political
barriers, to greater use of evidence in policymaking. We believe that many in
the policy community will respond positively.
C. SUMMARY
Our proposals, in some ways, amount to no less than a call for a culture shift
within the legal academy. We suggest that legal scholars transform themselves
from expert-advocates into scientists who approach inquiries through hypotheses, grounded in theory or commonly-held assumptions, rather than a particular position; who give equal weight to all available evidence, whether favorable
or unfavorable to the hypotheses; who report all empirical results, regardless of
whether they confirm or call into question their initial position; and who
affirmatively disclose the shortcomings of their work. We further suggest that
legal scholars should acquire additional expertise (either personally or through
collaborations) to accomplish this objective, even though their investment will
not be rewarded (at least initially) in the usual processes through which the legal
academy recognizes scholarly achievements. Despite the challenges rooted in
the different and sometimes conflicting methodologies of legal scholars and
MANUAL ON SCIENTIFIC EVIDENCE 179 (2d ed. 2000), available at http://www.fjc.gov/public/pdf.nsf/lookup/
11.mult_reg.pdf/$File/11.mult_reg.pdf (educating judges about regression analysis); Jeffrey C. Lerner,
Book Review, 28 J. LEGAL MED. 283, 290–91 (2007) (reviewing RESCUING SCIENCE FROM POLITICS:
REGULATION AND THE DISTORTION OF SCIENTIFIC RESEARCH (Wendy Wagner & Rena Steinzor eds., 2006),
which describes a program at Brooklyn Law School that educates judges on empirical methodology,
including systematic reviews).
252. Eisenberg, supra note 229.
702
THE GEORGETOWN LAW JOURNAL
[Vol. 96:649
social scientists,253 we can do much to bridge the gap and increase the impact of
empirical legal studies on policy.
Clearly, this is not for everyone, or even every law school. And indeed, it is
critical to preserve what is distinctive and valuable about traditional modes of
legal scholarship—for example, the depth of study and exposition, the consideration of political and moral dimensions of policy problems, and the normative
recommendations that point judges and policy makers in a clear direction. Our
claim is simply that empirical legal research should be afforded the same
respect and nurtured to the same extent as more traditional forms of inquiry.
This involves recognizing the unique challenges that empirical legal studies
face, as well as the special contributions it has made and can make toward
betterment of public policy.
CONCLUSION
The field of empirical health law is in its infancy, but it is growing quickly.
Several forces have aligned to support its continued expansion. Unless law
schools undergo substantial changes, however, this expansion will continue
mostly outside the legal academy with limited involvement of legal scholars,
who have the potential to bring a deep understanding of law and legal institutions to the enterprise. Empirical health law research stands to benefit greatly
from changes within the legal academy to foster an increase not only in the
production of this research but also in the impact it has on enactment of
effective legislation and dispute resolution. Despite the challenges we have
outlined, we remain optimistic that the field of empirical health law will
continue to grow and mature, and that the great public health achievements of
the twenty-first century, like many of those of the last century, will owe much to
its contributions.
253. See Blumenthal, supra note 227, at 47 (“(1) [S]ocial science is innovative, while law resists
innovation, (2) social science is based on data and observation, while law is based on precedent and
hierarchy, (3) social science seeks an objective answer to problems, while law seeks an adversarial
victory, (4) social science is descriptive, while law is prescriptive, (5) social science is nomothetic,
while law is idiographic, (6) social science conclusions are probabilistic and tentative, while legal
conclusions are irrevocable and must appear certain, (7) social science is proactive, while law is
reactive, and (8) social science is abstract, while law deals with concrete issues.” (quoting J. Alexander
Tanford, 66 IND. L.J. 137, 156 (1990))).